AtomMem: Building Simple and Effective Memory System for LLM Agents via Atomic Facts

arXiv:2606.19847 · cs.CL · Submitted 2026-06-18 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: AI Radio. Generated commentary on the latest Artificial Intelligence papers.

Tom: We're looking at the paper: "AtomMem: Building Simple and Effective Memory System for LLM Agents via Atomic Facts".

Jane: Here's the gist of the paper. Large language models (LLMs) face challenges with long-term information accumulation and reuse across multi-session interactions due to fixed context windows.

Tom: AtomMem is proposed as a long-term memory system designed for value-dense storage and stable memory evolution.

Jane: Core System Architecture AtomMem introduces a Fact Executor to selectively extract high value atomic facts from long form interactions to serve as highly efficient memory representations.

Tom: These facts are then organized into hierarchical event structures and temporal profiles to capture coherent episodic contexts and track dynamically evolving user attributes over time.

Title and authors: Tom: So, we're looking at this paper called AtomMem: Building Simple and Effective Memory System for LLM Agents via Atomic Facts. It seems like they're tackling the big problem with current LLM memory, which is that their context windows are too small for keeping track of a long conversation or multiple sessions.

Jane: That's right, and what they propose is this long-term memory system designed to store information in a way that’s actually useful and evolves over time instead of just getting messy.

Lu: What makes this interesting from my side is their core idea: they introduce a Fact Executor that selectively pulls out these high-value atomic facts from long interactions, which are then organized into hierarchical event structures and temporal profiles.

Meng: From a deployment standpoint, I'm curious about how they handle the sheer volume of data being extracted. If we're talking about real traffic, what does this atomic fact extraction process look like in terms of overhead?

Lalam: I think the core mechanism is that they treat every piece of information as an atomic fact F equals c, where c is just the self-contained text generated by a specialized LLM extractor.

Tom: Exactly. And they structure these facts with some metadata—id, c, value v, P for properties and keys and T for time and E for event—which makes them much more efficient to use later than just dumping raw text into a database.

Jane: The summary of the paper talks about how this organization lets them capture both episodic contexts from the conversation and track how user attributes evolve throughout those sessions using these temporal profiles.

Lu: And they also have this memory graph activation mechanism where they use an associative memory graph over those atomic facts to connect fragmented memories across different interactions.

Meng: Connecting them across sessions sounds powerful, but that's where I worry about stability. If you have a million facts scattered everywhere, how do you make sure the retrieval isn't just pulling random noise?

Tom: That’s where they introduce three types of associations as edges in their graph: an Entity Edge connecting facts with shared keywords, an Event Edge for facts belonging to the same event, and a Temporal Edge connecting those in adjacent dialogue turns.

Jane: And they use something called Random Walk with Restart, or RWR, to propagate activations across this graph to find relevant memories when you ask a question.

Lalam: The retrieval part is layered too; they have three stages: Primary Recall based on participant and temporal constraints, Compensatory Recall using event layer scoring, and then Associative Recall via that graph activation.

Title and authors: Meng: Compensatory recall sounds like a good safety net when the primary recall misses something, but how do you tune that fusion weight to get it right?

Tom: The paper suggests tuning this compensatory fusion weight 'w' to find an optimal peak at w equals zero point seven across all core evaluation dimensions.

Jane: It sounds like they're trying to balance finding the big picture event context with getting super precise on the individual facts.

Lu: They also have a feature called Profile Augmentation, which retrieves stable user attributes if the intent analysis sets Iprof to one, meaning it pulls in those long-term profile entries.

Meng: If we're deploying this at scale, what about updating these profiles? The paper mentions using an LLM-driven updater for state evolution during session batch processing. That sounds like a potential failure point.

Lalam: They specifically address that by strengthening the temporal profile modeling and ensuring that when a profile changes, AtomMem correctly copies previous states into history H and updates historical versions to preserve past user states.

Tom: So, the improvements they suggest are pretty detailed: enhancing fact extraction for better coreference and temporal anchoring, refining the Entity Edge weight calculation using something like w kw of Fi, Fj equals IDF(k) times beta q of k times pi(k), and improving episodic consolidation by making the LLM's routing decisions more robust.

Jane: And they also fine-tuned the state evolution stability by improving that LLM-driven updater logic during batch processing to make sure past states are preserved correctly.

Lu: Plus, for retrieval precision, they suggest optimizing the Compensatory Recall stage by tuning that fusion weight 'w' to hit an optimal peak of zero point seven.

Meng: From my view, these technical tweaks sound good on paper, but I still have to ask about the computational cost. The paper says their simplified fact-level variant shows competitive performance at minimal computational cost, but what about the full modular design?

Tom: The paper notes that the full modular design yields further significant gains over the simplified fact-level variant. And it does show they reduce total token consumption by approximately sixty point one four percent compared to Mem0.

Jane: That reduction in token usage is a big deal if we're thinking about how much data we can feed into these systems before hitting those context window limits.

Lu: The overall finding is that AtomMem consistently outperforms state-of-the-art baselines, and the simplified fact-level variant achieves competitive performance even with minimal computational cost.

Meng: So for production, it looks like a modular design offers more power, but the core value seems to be in how efficiently they represent and retrieve those atomic facts.

Title and authors: Lalam: I'm confident that this approach of organizing interactions into these structured events and using graph-based recall is what makes AtomMem a long-term memory system for LLM agents.

Tom: So, to wrap up on the implications, this paper shows how we can move beyond flat retrieval over isolated items and create a system centered on atomic facts that organizes user interactions into hierarchical structures.

Jane: It means AI agents can maintain a much richer history of personalized context across many different conversations and sessions.

Lu: It shifts the focus from just scaling the model size to building efficient, structured memory systems that handle long-horizon reasoning and complex associations.

Meng: For us, it means we have a clearer path on what kind of memory structure would actually be stable enough for real-world deployment without constant catastrophic forgetting.

Lalam: The paper AtomMem: Building Simple and Effective Memory System for LLM Agents via Atomic Facts lays out a solid foundation for building these long-term memory systems.

Tom: Yeah, it’s a system that aims to be value-dense and stable in its evolution.

Jane: It really shows how structuring the facts—the c, v, P, K, T, E—allows the system to do more than just store text; it enables complex reasoning across time.

Lu: I think this paper is important because it tackles the stability issue head-on by proposing those mechanisms for stable and controllable memory updates.

Meng: It’s interesting how they are using graph structures to connect facts that might have been separated by different sessions, which is a key challenge we face in agent design.

Lalam: And the three types of edges—Entity, Event, and Temporal—give the system a very specific way to navigate and recall information based on what kind of connection you're looking for.

Tom: So, we’ve seen how they organize these facts into events and profiles, how they use graph activation to find connections, and how they tune the retrieval stages for precision.

Jane: Overall, AtomMem seems like a very concrete system designed to give LLM agents a persistent understanding of their own history.

Lu: It’s not just about getting better at retrieving things; it's about building memory that is organized hierarchically and can adapt to new information in a structured way.

Meng: I think the reduction in token consumption, even by sixty point one four percent compared to Mem0, makes this look like a practical system for running on existing infrastructure.

Lalam: The paper AtomMem: Building Simple and Effective Memory System for LLM Agents via Atomic Facts provides a clear roadmap for building memory systems that are structured, value-dense, and evolve stably.

The paper's summary: Tom: So, AtomMem is basically taking all that messy conversation history and boiling it down into these super clean, self-contained pieces of information called atomic facts.

Jane: That’s right, Tom, they’re not just storing the text; they’re extracting high-value facts—like a specific user preference or an event that happened—and then turning those into structured data points.

Lu: The core idea is that instead of trying to manage one giant context window for everything, you break it down into these events and profiles, which lets the memory evolve over time in a way that makes sense.

Meng: I’m thinking about the practical side here; if this system can handle millions of facts across many sessions, how much overhead is that on the actual hardware?

Lalam: The system uses a lot of organization—hierarchical events and temporal profiles—to capture both what happened in an event and who the user was before or after.

Tom: And to make retrieval work across all those different pieces of information, they build this associative memory graph, connecting facts based on shared words, shared events, or just being close in time.

Jane: That graph activation is how the system finds related memories when you ask a question; it’s like tracing a path through the history to find the right clue.

Lu: The retrieval process itself has three layers: one based on what you are asking and when, another layer that looks at the event layers, and then that graph recall we talked about before.

Tom: And they include this profile augmentation step if the intent analysis points toward needing stable user attributes, which is important because those profiles track how a person changes.

Meng: So it’s designed to be robust enough for complex reasoning tasks, but I still wonder about the actual computational cost of running all that graph traversal every time.

Jane: The interesting part is that they show this system performing at state-of-the-art levels on tests like LoCoMo, and their simplified version shows it’s competitive even when you reduce the complexity.

Lu: And they specifically reduced total token consumption by sixty point one four percent compared to a system called Mem0, which is a decent efficiency gain.

Tom: That token saving is significant because it means less input data goes into the model's context window every time it runs something.

Jane: What this means for us in terms of real-world agents is that we can have these agents maintain a much richer, long-term understanding of their user without hitting those hard limits too quickly.

Lu: It shifts the focus away from just making the base language model bigger and toward building these better structured memory systems that handle long-horizon reasoning.

Meng: For me, it’s about deploying something that doesn't break at scale, and if you can manage that fact extraction efficiently, it looks like a system we could actually put into production.

Lalam: I’m confident because the paper shows they didn't just build a big database; they built this way to organize facts so they are useful for reasoning.

Tom: Exactly, it’s about moving from flat retrieval of isolated text to a system that understands the relationship between all those stored pieces.

Jane: It gives these agents a persistent memory, not just for what happened in the last five minutes of chat, but across weeks or even months of interaction.

Lu: The main thing they nail is making sure that when user attributes update, like a preference shifts, the system correctly preserves the old state while adding history.

Meng: It sounds like it addresses a lot of the stability issues we see with memory systems that just keep stacking new data without cleaning up or versioning properly.

Tom: So, to recap: AtomMem uses atomic facts and an associative graph to build a stable, evolving long-term memory structure for LLM agents.

Jane: It’s about making sure the AI doesn't forget who it was talking to or what happened earlier in the conversation when things get complicated.

The paper's improvements: Tom: So, the authors aren't just happy with the base design; they actually suggest several ways to make AtomMem better and more stable in practice.

Jane: They are focusing on three main areas of improvement: how they extract those initial facts, how they build the memory graph, and how they manage state changes over time.

Lu: First, on fact extraction, they want to improve the coreference resolution so that every atomic fact generated is truly independent and understandable on its own.

Meng: That sounds like a big win for deployment because if the facts are clean, it means less noise when we feed them into the retrieval system.

Tom: Second, they suggest refining how those Entity Edges in the memory graph get their weights; they want to use a more complex function involving inverse document frequency and keyword probability to better balance semantic density against just having too many common words.

Jane: That’s smart because it prevents the system from getting bogged down by terms that don't actually tell you anything important.

Lu: Then, on episodic consolidation, they want the routing decisions made by the LLM when deciding which events to absorb or which new facts to trigger be more robust so it handles really diverse contexts better.

Meng: That robustness is what I need; if the AI gets confused about whether a new piece of information belongs in an old event or starts a brand new one, it breaks at scale.

Tom: And finally, for state evolution stability, they want to strengthen the logic in that LLM-driven updater during batch processing to make sure that when a profile changes, AtomMem correctly copies the old state into history and updates those older versions properly.

Jane: That preservation of past user states is crucial because if an AI forgets a preference it had last week, that’s a major failure in any personalized agent.

Lu: They also tweaked the retrieval precision stage by suggesting tuning that compensatory fusion weight to hit an optimal peak at zero point seven across all core evaluation dimensions.

Meng: Tuning parameters like that helps reduce the reliance on just one perfect setting and makes the system more resilient when things aren't perfectly structured.

Tom: The implication is that this isn't just a static memory; it’s a dynamic system where the AI actively learns how to manage its own history structure to keep things organized.

Jane: It means we can expect agents to maintain a much richer, more personalized context over long periods, which is what people really want from these systems.

Lu: The authors also pointed out that the current framework only processes textual interactions, so extending it to handle multimodal inputs is the natural next step for this research direction.

Conclusion: Tom: So we’ve gone through how AtomMem takes messy chat history and turns it into structured atomic facts for agents to use as long-term memory.

Jane: Exactly, and the big picture is that this moves us closer to having AI agents that actually remember things across sessions without constantly forgetting them.

Lu: The implication here is a system that can handle long-horizon reasoning much more reliably because the memory isn't just a bag of text; it's organized hierarchically.

Meng: For deployment, if we can get that structured representation with the token savings they showed, it looks like a much more sustainable way to build agents at scale.

Lalam: I’m confident this paper delivers on its promise because the system is designed specifically to organize facts in a way that makes them useful for reasoning across different interactions.

Tom: It’s about building value-dense memory, not just volume, which is what really matters when you’re trying to keep an agent coherent over time.

Jane: It shows how structuring the facts—the text plus the metadata like time and event—allows the AI to do much more than just store text; it enables complex reasoning across different times.

Lu: They’ve also laid out clear paths for future work, pointing out that extending this framework to multimodal inputs is a natural next step for this research direction.

Meng: I’m still focused on the cost; if the fact extraction module gets too complex, we might run into issues keeping the memory retrieval fast enough for real-time use.

Tom: We’ll need to see how those optimization suggestions play out when we put them into a production environment, but this foundation is solid.

Jane: It really shows how structuring the facts—the c, v, P, K, T, E—allows the system to do more than just store text; it enables complex reasoning across different times.

Lu: The main thing they nail is making sure that when user attributes update, like a preference shifts, the system correctly preserves the old state while adding history.

Lalam: I’m confident because the paper shows they didn't just build a big database; they built this way to organize facts so they are useful for reasoning across different interactions.

Tom: So we’ve seen how they organize these facts into events and profiles, how they use graph activation to find connections, and how we tune the retrieval stages for precision.

Jane: Overall, AtomMem seems like a very concrete system designed to give AI agents a persistent understanding of their own history.

Lu: It’s not just about getting better at retrieving things; it's about building memory that is organized hierarchically and can adapt to new information in a structured way.

Meng: The token consumption reduction, even by sixty point one four percent compared to Mem0, makes this look like a practical system for running on existing infrastructure.

Tom: Yeah, it’s a system that aims to be value-dense and stable in its evolution.

Jane: It really shows how structuring the facts—the c, v, P, K, T, E—allows the system to do more than just store text; it enables complex reasoning across different times.

Lu: I think this paper is important because it tackles the stability issue head-on by proposing those mechanisms for stable and controllable memory updates.

Lalam: The paper AtomMem: Building Simple and Effective Memory System for LLM Agents via Atomic Facts provides a clear roadmap for building memory systems that are structured, value-dense, and evolve stably.

Tom: We’ve seen how they organize these facts into events and profiles, how they use graph activation to find connections, and how we tune the retrieval stages for precision.

Yanyu Yao, Shangze Li, Zhi Zheng, Hui Zheng, Qi Liu, Tong Xu, Enhong Chen

State Key Laboratory of Cognitive Intelligence, University of Science and Technology of China · Anhui University

cs.CL

Submitted: 2026-06-18

Updated: 2026-10-31

Comments: 23 pages, 10 figures, 8 tables

Code: https://github.com/MINE-USTC/AtomMem

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 88/100

The gist: Large language models (LLMs) face challenges with long-term information accumulation and reuse across multi-session interactions due to fixed context windows.

Key concepts

Atomic Fact Representation
The basic unit of memory is an atomic fact (F = id, c, v, P, K, T, E). The text 'c' is a self-contained piece of information generated by a specialized LLM. This ensures every stored piece of data is independent and understandable on its own without needing the original conversation context.
Memory Graph Association
AtomMem builds a graph connecting atomic facts using three types of edges: Entity, Event, and Temporal. These connections allow the system to find related memories even if they are far apart in time or conversation flow. A Random Walk with Restart (RWR) is used to activate these connections and link fragmented memories together.
Temporal Profile Modeling
Profiles capture stable user attributes over time, structured as P = u, c, vp, Kp, Eevi, tfrom. When a user's state changes during processing batches, the system uses an LLM-driven updater to correctly copy the previous state into history (H) and update historical versions. This preserves the past evolution of user attributes.
Hierarchical Retrieval Strategy
Retrieval happens in three stages: Primary Recall (based on time/participant), Compensatory Recall (using event scoring), and Associative Recall (using the memory graph). If intent analysis requires it, Profile Augmentation retrieves stable user attributes to enrich the final answer.

Terminology

Summary

Large language models (LLMs) face challenges with long-term information accumulation and reuse across multi-session interactions due to fixed context windows. p. 0

Where the paper says it

cn, huizheng@ahu. cn, Abstract Large language models (LLMs) demonstrate strong reasoning and generation abilities, but their fixed context windows limit long-term information accumulation and reuse across multi-session interactions.

AtomMem is proposed as a long-term memory system designed for value-dense storage and stable memory evolution. p. 0
Where the paper says it

Existing memoryaugmented systems often construct memory in a coarse and unstable manner, relying on inefficient memory representations or unstable unconstrained updates. To address these challenges, we propose AtomMem, a long-term memory system designed for value-dense storage and stable memory evolution.

Core System Architecture

AtomMem introduces a Fact Executor to selectively extract high value atomic facts from long form interactions to serve as highly efficient memory representations. These facts are then organized into hierarchical event structures and temporal profiles to capture coherent episodic contexts and track dynamically evolving user attributes over time. p. 0

Where the paper says it

AtomMem introduces a Fact Executor, which selectively extracts high value atomic facts from long form interactions to serve as highly efficient memory representations. Subsequently, AtomMem organizes these facts into hierarchical event structures and temporal profiles, capturing coherent episodic contexts and tracking dynamically evolving user attributes over time.

During retrieval, the system activates an associative memory graph to connect fragmented memories. p. 0
Where the paper says it

Subsequently, AtomMem organizes these facts into hierarchical event structures and temporal profiles, capturing coherent episodic contexts and tracking dynamically evolving user attributes over time. During retrieval, the system activates an associative memory graph to connect fragmented memories.

Atomic Fact Representation

The system's base representation is the atomic fact F = c where id is the fact identifier and c is the self-contained text generated by the extractor. This extraction module uses an SFT-tuned LLM to perform essential denoising and lightweight reasoning such as coreference resolution. p. 2

Where the paper says it

Therefore, our goal is to transform raw dialogue sessions into a set of structured, self-contained atomic facts F, thereby providing a superior memory representation. 1 Atomic Fact Extractor To address redundancy and noise, we introduce an Atomic Fact Extractor trained via supervised fine-tuning (SFT) that performs essential denoising and lightweight reasoning such as coreference resolution.

Structured metadata is added by defining F as F = id, c, v, P, K, T, E where v denotes its dense semantic embedding. p. 3
Where the paper says it

Formally, we define an atomic fact F as: F = {id, c, v,P, K, T , E} where id is the fact identifier, c is the self-contained text generated by the extractor, and v denotes its dense semantic embedding. To obtain precise symbolic metadata, we leverage the LLM to parse the conversation and extract the following key attributes.

Episodic and Temporal Structuring

Memory is structured into Events E = S where S is the concise summary and Fids is the set of constituent Fact ids. A profile entry P is structured as P = u, c, vp, Kp, Eevi, tfrom, H to capture stable attributes and historical versions. p. 4

Where the paper says it

Therefore, we introduce a profile layer to capture these persistent yet dynamic user states. Formally, a profile entry P is structured as follows: P = {id, u, c, vp, Kp, Eevi, tfrom, H} where u denotes the user, c describes the stable attribute of user, vp represents the vector embedding of the content, Kp is the keyword set, tfrom is the effective timestamp, H stores historical versions, and Eevi tracks supporting fact ids for traceability.

State evolution uses a session-based batch mechanism to process queued candidate facts into structured profiles. p. 4
Where the paper says it

During atomic fact extraction, the LLM identifies facts implying potential long-term characteristics and temporarily adds them to a waiting queue. At the end of a dialogue session, the system processes these queued candidate facts in a batch to generate structured profiles.

Associative Recall Mechanism

AtomMem activates a memory graph over atomic facts using three distinct types of associations as edges. These edges include an Entity Edge connecting two facts if they share keywords, an Event Edge connecting facts belonging to the same event, and a Temporal Edge connecting facts in adjacent dialogue turns. p. 4

Where the paper says it

This graph uses atomic facts as nodes and encodes three distinct types of associations as edges. Entity Edge Two facts connect if they share keywords.

The graph activation uses Random Walk with Restart (RWR) to propagate activations across these edges. p. 5
Where the paper says it

AtomMem then applies Random Walk with Restart (RWR) to propagate activations across entity, event and temporal edges. Finally, the system selects the top kf facts with the highest activation scores across all nodes to output the final retrieved context Rfact.

Hierarchical Retrieval and Response

Retrieval involves a three-stage hierarchical retrieval strategy: Primary Recall based on participant and temporal constraints, Compensatory Recall using event layer scoring, and Associative Recall via graph activation. Profile Augmentation retrieves stable user attributes if the intent analysis sets Iprof = 1. p. 5

Where the paper says it

Finally, the system selects the top kf facts with the highest activation scores across all nodes to output the final retrieved context Rfact. 3 Profile Augmentation and Response Generation If the intent analysis sets the flag Iprof = 1, AtomMem executes profile recall to retrieve stable user attributes.

The final response is constructed by concatenating the retrieved episodic fact set Rfact and the semantic profile set Rprof. p. 5
Where the paper says it

Finally, AtomMem constructs a comprehensive context C by concatenating the retrieved episodic fact set Rfact and the semantic profile set Rprof . This context C, along with the original query q, is fed into the LLM to generate the final response.

Performance and Efficiency

Experiments on the LoCoMo benchmark confirm that AtomMem achieves state-of-the-art performance across various reasoning tasks. The simplified fact-level variant demonstrates competitive performance at minimal computational cost. p. 2

Where the paper says it

benchmarks demonstrate that AtomMem consistently outperforms state-of-the-art baselines. Notably, our simplified fact-level variant achieves competitive performance at minimal computational cost, while the full modular design yields further significant gains.

The system reduces total token consumption by approximately 61. p. 6
Where the paper says it

Computational Efficiency AtomMem demonstrates competitive cost efficiency alongside its robust reasoning capabilities. Despite the overhead associated with memory construction, our system significantly reduces total token consumption by approximately 61.

4% compared to the highly performant Mem0. p. 6
Where the paper says it

Despite the overhead associated with memory construction, our system significantly reduces total token consumption by approximately 61. 4% compared to the highly performant Mem0.

The end-to-end latency averages approximately 3. p. 4
Where the paper says it

During atomic fact extraction, the LLM identifies facts implying potential long-term characteristics and temporarily adds them to a waiting queue. At the end of a dialogue session, the system processes these queued candidate facts in a batch to generate structured profiles.

6 seconds across all processing stages. p. 5
Where the paper says it

We employ an LLM to parse q and extract a structured query object Qparsed, defined as: Qparsed = {Iprof ,Pq, Kq, Tq} where Iprof ∈ {0, 1} indicates whether user profiles are required, Pq denotes the involved participants, Kq captures the core intent via extracted keywords, and Tq specifies the relevant time range. 2 Hierarchical Hybrid Retrieval To balance precision and contextual breadth, we implement a hierarchical retrieval strategy consisting of three stages.

The optimal configuration for the retrieval pipeline is detailed in Table 5. p. 6
Where the paper says it

Detailed hyperparameter configurations for AtomMem’s hierarchical retrieval and fusion mechanisms are provided in Appendix A. 2 Main Results Superior Performance As presented in Table 1, AtomMem achieves state-of-the-art results across all evaluated metrics.

Limitations

A few limitations include reliance on the capabilities of the underlying LLM for performance sensitivity. The current framework processes only textual interactions, and extension to multimodal inputs is a natural progression. p. 8

Where the paper says it

Second, the current framework processes only textual interactions. Real world conversations often involve multimodal inputs like images and audio, making system extension a natural progression for future research.

Token efficiency can be further optimized in the future. p. 8
Where the paper says it

Real world conversations often involve multimodal inputs like images and audio, making system extension a natural progression for future research. Lastly, although our system achieves a favorable balance between token efficiency and system performance, token efficiency can be further optimized.

References

Akari Asai, Zeqiu Wu, Yizhong Wang, Avirup Sil, and Hannaneh Hajishirzi. 2024. Self-RAG: Learning to retrieve, generate, and critique through self-reflection. p. 8

Where the paper says it

References Akari Asai, Zeqiu Wu, Yizhong Wang, Avirup Sil, and Hannaneh Hajishirzi. Self-RAG: Learning to retrieve, generate, and critique through self-reflection.

In The Twelfth International Conference on Learning Representations. p. 8
Where the paper says it

Self-RAG: Learning to retrieve, generate, and critique through self-reflection. In The Twelfth International Conference on Learning Representations.

p. 8
Where the paper says it

Self-RAG: Learning to retrieve, generate, and critique through self-reflection. In The Twelfth International Conference on Learning Representations.

Sébastien Bubeck, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee, Yin Tat Lee, Yuanzhi Li, Scott Lundberg, and 1 others. p. 8

Where the paper says it

In The Twelfth International Conference on Learning Representations. Sébastien Bubeck, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee, Yin Tat Lee, Yuanzhi Li, Scott Lundberg, and 1 others.

p. 8
Where the paper says it

In The Twelfth International Conference on Learning Representations. Sébastien Bubeck, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee, Yin Tat Lee, Yuanzhi Li, Scott Lundberg, and 1 others.

p. 8
Where the paper says it

In The Twelfth International Conference on Learning Representations. Sébastien Bubeck, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee, Yin Tat Lee, Yuanzhi Li, Scott Lundberg, and 1 others.

2023. Sparks of artificial general intelligence: Early experiments with GPT-4. p. 8
Where the paper says it

Sébastien Bubeck, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee, Yin Tat Lee, Yuanzhi Li, Scott Lundberg, and 1 others. Sparks of artificial general intelligence: Early experiments with GPT-4.

Preprint.

Prateek Chhikara, Dev Khant, Saket Aryan, Taranjeet Singh, and Deshraj Yadav. p. 8

Where the paper says it

Preprint, arXiv:2303. Prateek Chhikara, Dev Khant, Saket Aryan, Taranjeet Singh, and Deshraj Yadav.

2025. Mem0: Building production-ready AI agents with scalable long-term memory. p. 8
Where the paper says it

Prateek Chhikara, Dev Khant, Saket Aryan, Taranjeet Singh, and Deshraj Yadav. Mem0: Building production-ready AI agents with scalable long-term memory.

Preprint.

Jizhan Fang, Xinle Deng, Haoming Xu, Ziyan Jiang, Yuqi Tang, Ziwen Xu, Shumin Deng, Yunzhi Yao, Mengru Wang, Shuofei Qiao, Huajun Chen, and Ningyu Zhang. p. 8

Where the paper says it

Preprint, arXiv:2504. Jizhan Fang, Xinle Deng, Haoming Xu, Ziyan Jiang, Yuqi Tang, Ziwen Xu, Shumin Deng, Yunzhi Yao, Mengru Wang, Shuofei Qiao, Huajun Chen, and Ningyu Zhang.

2026. LightMem: Lightweight and efficient memory-augmented generation. p. 8
Where the paper says it

Jizhan Fang, Xinle Deng, Haoming Xu, Ziyan Jiang, Yuqi Tang, Ziwen Xu, Shumin Deng, Yunzhi Yao, Mengru Wang, Shuofei Qiao, Huajun Chen, and Ningyu Zhang. LightMem: Lightweight and efficient memory-augmented generation.

In The Fourteenth International Conference on Learning Representations. p. 8
Where the paper says it

LightMem: Lightweight and efficient memory-augmented generation. In The Fourteenth International Conference on Learning Representations.

Yunfan Gao, Yun Xiong, Xinyu Gao, Kangxiang Jia, Jinliu Pan, Yuxi Bi, Yi Dai, Jiawei Sun, Meng Wang, and Haofen Wang. p. 8

Where the paper says it

In The Fourteenth International Conference on Learning Representations. Yunfan Gao, Yun Xiong, Xinyu Gao, Kangxiang Jia, Jinliu Pan, Yuxi Bi, Yi Dai, Jiawei Sun, Meng Wang, and Haofen Wang.

2023. Retrieval-augmented generation for large language models: A survey. p. 8
Where the paper says it

Yunfan Gao, Yun Xiong, Xinyu Gao, Kangxiang Jia, Jinliu Pan, Yuxi Bi, Yi Dai, Jiawei Sun, Meng Wang, and Haofen Wang. Retrieval-augmented generation for large language models: A survey.

Preprint.

Kelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat, and Mingwei Chang. p. 8

Where the paper says it

Preprint, arXiv:2312. Kelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat, and Mingwei Chang.

2020. Retrieval augmented language model pre-training. p. 8
Where the paper says it

Kelvin Guu, Kenton Lee, Zora Tung, Panupong Pasupat, and Mingwei Chang. Retrieval augmented language model pre-training.

In Proceedings of the 37th International Conference on Machine Learning. p. 8
Where the paper says it

Retrieval augmented language model pre-training. In Proceedings of the 37th International Conference on Machine Learning, volume 119 of Proceedings of Machine Learning Research, pages 3929–3938.

Zhengbao Jiang, Frank Xu, Luyu Gao, Zhiqing Sun, Qian Liu, Jane Dwivedi-Yu, Yiming Yang, Jamie Callan, and Graham Neubig. p. 8

Where the paper says it

In Proceedings of the 37th International Conference on Machine Learning, volume 119 of Proceedings of Machine Learning Research, pages 3929–3938. Zhengbao Jiang, Frank Xu, Luyu Gao, Zhiqing Sun, Qian Liu, Jane Dwivedi-Yu, Yiming Yang, Jamie Callan, and Graham Neubig.

2023. Active retrieval augmented generation. p. 8
Where the paper says it

Zhengbao Jiang, Frank Xu, Luyu Gao, Zhiqing Sun, Qian Liu, Jane Dwivedi-Yu, Yiming Yang, Jamie Callan, and Graham Neubig. Active retrieval augmented generation.

In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing. p. 8
Where the paper says it

Active retrieval augmented generation. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 7969–7992, Singapore.

Jiazheng Kang, Mingming Ji, Zhe Zhao, and Ting Bai. p. 8

Where the paper says it

Association for Computational Linguistics. Jiazheng Kang, Mingming Ji, Zhe Zhao, and Ting Bai.

2025. Memory OS of AI agent. p. 2
Where the paper says it

, 2024) further introduced dynamic retrieval timing and output critiquing. In LLM agents, retrieval-based access has become an important mechanism for exposing external knowledge and long-term memory to the agent (Park et al.

In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. p. 8
Where the paper says it

Memory OS of AI agent. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 25961–25970, Suzhou, China.

Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel. p. 8

Where the paper says it

Association for Computational Linguistics. Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel, and Douwe Kiela. p. 8

Where the paper says it

Association for Computational Linguistics. Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel, and Douwe Kiela.

, and Douwe Kiela. 2020. Retrieval-augmented generation for knowledgeintensive NLP tasks. p. 8
Where the paper says it

Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel, and Douwe Kiela. Retrieval-augmented generation for knowledgeintensive NLP tasks.

In Advances in Neural Information Processing Systems. p. 8
Where the paper says it

Retrieval-augmented generation for knowledgeintensive NLP tasks. In Advances in Neural Information Processing Systems, volume 33, pages 9459– 9474.

p. 8
Where the paper says it

Retrieval-augmented generation for knowledgeintensive NLP tasks. In Advances in Neural Information Processing Systems, volume 33, pages 9459– 9474.

Lei Liu, Xiaoyan Yang, Yue Shen, Binbin Hu, Zhiqiang Zhang, Jinjie Gu, and Guannan Zhang. p. 8

Where the paper says it

In Advances in Neural Information Processing Systems, volume 33, pages 9459– 9474. Lei Liu, Xiaoyan Yang, Yue Shen, Binbin Hu, Zhiqiang Zhang, Jinjie Gu, and Guannan Zhang.

2023. Think-in-memory: Recalling and post-thinking enable LLMs with long-term memory. p. 8
Where the paper says it

Lei Liu, Xiaoyan Yang, Yue Shen, Binbin Hu, Zhiqiang Zhang, Jinjie Gu, and Guannan Zhang. Think-in-memory: Recalling and post-thinking enable LLMs with long-term memory.

Preprint.

Nelson F. Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang. p. 8

Where the paper says it

Preprint, arXiv:2311. Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang.

2024. Lost in the middle: How language models use long contexts. p. 8
Where the paper says it

Liu, Kevin Lin, John Hewitt, Ashwin Paranjape, Michele Bevilacqua, Fabio Petroni, and Percy Liang. Lost in the middle: How language models use long contexts.

Transactions of the Association for Computational Linguistics. p. 8
Where the paper says it

Lost in the middle: How language models use long contexts. Transactions of the Association for Computational Linguistics, 12:157–173.

Shuochen Liu, Junyi Zhu, Long Shu, Junda Lin, Yuhao Chen, Haotian Zhang, Chao Zhang, Derong Xu, Jia Li, Bo Tang, Zhiyu Li. p. 8

Where the paper says it

Transactions of the Association for Computational Linguistics, 12:157–173. Shuochen Liu, Junyi Zhu, Long Shu, Junda Lin, Yuhao Chen, Haotian Zhang, Chao Zhang, Derong Xu, Jia Li, Bo Tang, Zhiyu Li, Feiyu Xiong, Enhong Chen, and Tong Xu. p. 0

Where the paper says it

AtomMem: Building Simple and Effective Memory System for LLM Agents via Atomic Facts Yanyu Yao1 , Shangze Li1 , Zhi Zheng1 , Hui Zheng2 , Qi Liu1 , Tong Xu1 , Enhong Chen1 1State Key Laboratory of Cognitive Intelligence, University of Science and Technology of China, Hefei, China 2Anhui University, Hefei, China {yyyao, lishangze, liuqilq}@mail. cn, {zhengzhi97, tongxu, cheneh}@ustc.

, Feiyu Xiong. p. 8
Where the paper says it

Transactions of the Association for Computational Linguistics, 12:157–173. Shuochen Liu, Junyi Zhu, Long Shu, Junda Lin, Yuhao Chen, Haotian Zhang, Chao Zhang, Derong Xu, Jia Li, Bo Tang, Zhiyu Li, Feiyu Xiong, Enhong Chen, and Tong Xu.

, Enhong Chen. p. 8
Where the paper says it

Transactions of the Association for Computational Linguistics, 12:157–173. Shuochen Liu, Junyi Zhu, Long Shu, Junda Lin, Yuhao Chen, Haotian Zhang, Chao Zhang, Derong Xu, Jia Li, Bo Tang, Zhiyu Li, Feiyu Xiong, Enhong Chen, and Tong Xu.

, and Tong Xu. 2026. PERMA: Benchmarking personalized memory agents via event-driven preference and realistic task environments. p. 8
Where the paper says it

Shuochen Liu, Junyi Zhu, Long Shu, Junda Lin, Yuhao Chen, Haotian Zhang, Chao Zhang, Derong Xu, Jia Li, Bo Tang, Zhiyu Li, Feiyu Xiong, Enhong Chen, and Tong Xu. PERMA: Benchmarking personalized memory agents via event-driven preference and realistic task environments.

Preprint.

Adyasha Maharana, Dong-Ho Lee, Sergey Tulyakov, Mohit Bansal, Francesco Barbieri, and Yuwei Fang. p. 8

Where the paper says it

Preprint, arXiv:2603. Adyasha Maharana, Dong-Ho Lee, Sergey Tulyakov, Mohit Bansal, Francesco Barbieri, and Yuwei Fang.

2024. Evaluating very long-term conversational memory of LLM agents. p. 8
Where the paper says it

Adyasha Maharana, Dong-Ho Lee, Sergey Tulyakov, Mohit Bansal, Francesco Barbieri, and Yuwei Fang. Evaluating very long-term conversational memory of LLM agents.

In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). p. 8
Where the paper says it

Evaluating very long-term conversational memory of LLM agents. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 13851– 13870, Bangkok, Thailand.

Ali Modarressi, Ayyoob Imani, Mohsen Fayyaz, and Hinrich Schütze. p. 8

Where the paper says it

Association for Computational Linguistics. Ali Modarressi, Ayyoob Imani, Mohsen Fayyaz, and Hinrich Schütze.

2023. RET-LLM: Towards a general read-write memory for large language models. p. 8
Where the paper says it

Ali Modarressi, Ayyoob Imani, Mohsen Fayyaz, and Hinrich Schütze. RET-LLM: Towards a general read-write memory for large language models.

Preprint.

Rodrigo Nogueira and Kyunghyun Cho. p. 8

Where the paper says it

Preprint, arXiv:2305. Rodrigo Nogueira and Kyunghyun Cho.

2019. Passage re-ranking with BERT. p. 8
Where the paper says it

Rodrigo Nogueira and Kyunghyun Cho. Passage re-ranking with BERT.

Preprint.

OpenAI. 2023. GPT-4 technical report. p. 8

Where the paper says it

Preprint, arXiv:1901. GPT-4 technical report.

Preprint.

Charles Packer, Sarah Wooders, Kevin Lin, Vivian Fang, Shishir G. p. 8

Where the paper says it

Preprint, arXiv:2303. Charles Packer, Sarah Wooders, Kevin Lin, Vivian Fang, Shishir G.

Patil, Ion Stoica, and Joseph E. p. 8
Where the paper says it

Charles Packer, Sarah Wooders, Kevin Lin, Vivian Fang, Shishir G. Patil, Ion Stoica, and Joseph E.

Gonzalez. 2023. MemGPT: Towards LLMs as operating systems. p. 8
Where the paper says it

Patil, Ion Stoica, and Joseph E. MemGPT: Towards LLMs as operating systems.

Preprint.

Joon Sung Park, Joseph C. p. 9

Where the paper says it

Joon Sung Park, Joseph C. O’Brien, Carrie J. p. 9

Where the paper says it

Joon Sung Park, Joseph C. O’Brien, Carrie J.

O’Brien, Carrie J. Cai, Meredith Ringel Morris, Percy Liang. p. 9
Where the paper says it

O’Brien, Carrie J. Cai, Meredith Ringel Morris, Percy Liang, and Michael S. p. 9

Where the paper says it

O’Brien, Carrie J. Cai, Meredith Ringel Morris, Percy Liang, and Michael S.

, and Michael S. Bernstein. 2023. Generative agents: Interactive simulacra of human behavior. p. 9
Where the paper says it

Cai, Meredith Ringel Morris, Percy Liang, and Michael S. Generative agents: Interactive simulacra of human behavior.

In Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology. p. 9
Where the paper says it

Generative agents: Interactive simulacra of human behavior. In Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology, pages 1–22.

Zhen Tan, Jun Yan, I-Hung Hsu, Rujun Han, Zifeng Wang, Long Le, Yiwen Song, Yanfei Chen. p. 9

Where the paper says it

Association for Computing Machinery. Zhen Tan, Jun Yan, I-Hung Hsu, Rujun Han, Zifeng Wang, Long Le, Yiwen Song, Yanfei Chen, Hamid Palangi, George Lee, Anand Rajan Iyer, Tianlong Chen, Huan Liu, Chen-Yu Lee, and Tomas Pfister. p. 9

Where the paper says it

Association for Computing Machinery. Zhen Tan, Jun Yan, I-Hung Hsu, Rujun Han, Zifeng Wang, Long Le, Yiwen Song, Yanfei Chen, Hamid Palangi, George Lee, Anand Rajan Iyer, Tianlong Chen, Huan Liu, Chen-Yu Lee, and Tomas Pfister.

, Hamid Palangi, George Lee. p. 9
Where the paper says it

Association for Computing Machinery. Zhen Tan, Jun Yan, I-Hung Hsu, Rujun Han, Zifeng Wang, Long Le, Yiwen Song, Yanfei Chen, Hamid Palangi, George Lee, Anand Rajan Iyer, Tianlong Chen, Huan Liu, Chen-Yu Lee, and Tomas Pfister.

, Anand Rajan Iyer. p. 9
Where the paper says it

Association for Computing Machinery. Zhen Tan, Jun Yan, I-Hung Hsu, Rujun Han, Zifeng Wang, Long Le, Yiwen Song, Yanfei Chen, Hamid Palangi, George Lee, Anand Rajan Iyer, Tianlong Chen, Huan Liu, Chen-Yu Lee, and Tomas Pfister.

, Tianlong Chen. p. 9
Where the paper says it

Association for Computing Machinery. Zhen Tan, Jun Yan, I-Hung Hsu, Rujun Han, Zifeng Wang, Long Le, Yiwen Song, Yanfei Chen, Hamid Palangi, George Lee, Anand Rajan Iyer, Tianlong Chen, Huan Liu, Chen-Yu Lee, and Tomas Pfister.

, Huan Liu. p. 9
Where the paper says it

Association for Computing Machinery. Zhen Tan, Jun Yan, I-Hung Hsu, Rujun Han, Zifeng Wang, Long Le, Yiwen Song, Yanfei Chen, Hamid Palangi, George Lee, Anand Rajan Iyer, Tianlong Chen, Huan Liu, Chen-Yu Lee, and Tomas Pfister.

, Chen-Yu Lee. p. 9
Where the paper says it

Association for Computing Machinery. Zhen Tan, Jun Yan, I-Hung Hsu, Rujun Han, Zifeng Wang, Long Le, Yiwen Song, Yanfei Chen, Hamid Palangi, George Lee, Anand Rajan Iyer, Tianlong Chen, Huan Liu, Chen-Yu Lee, and Tomas Pfister.

, and Tomas Pfister. 2025. In prospect and retrospect: Reflective memory management for long-term personalized dialogue agents. p. 9
Where the paper says it

Zhen Tan, Jun Yan, I-Hung Hsu, Rujun Han, Zifeng Wang, Long Le, Yiwen Song, Yanfei Chen, Hamid Palangi, George Lee, Anand Rajan Iyer, Tianlong Chen, Huan Liu, Chen-Yu Lee, and Tomas Pfister. In prospect and retrospect: Reflective memory management for long-term personalized dialogue agents.

In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). p. 9
Where the paper says it

In prospect and retrospect: Reflective memory management for long-term personalized dialogue agents. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 8416–8439, Vienna, Austria.

Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava. p. 9

Where the paper says it

Association for Computational Linguistics. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, and 1 others.

, Shruti Bhosale. p. 9
Where the paper says it

Association for Computational Linguistics. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, and 1 others.

, and 1 others. 2023. Llama 2: Open foundation and fine-tuned chat models. p. 9
Where the paper says it

Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, and 1 others. Llama 2: Open foundation and fine-tuned chat models.

Preprint.

Di Wu, Hongwei Wang, Wenhao Yu, Yuwei Zhang, KaiWei Chang. p. 9

Where the paper says it

Preprint, arXiv:2307. Di Wu, Hongwei Wang, Wenhao Yu, Yuwei Zhang, KaiWei Chang, and Dong Yu. p. 8

Where the paper says it

Preprint, arXiv:2603. Adyasha Maharana, Dong-Ho Lee, Sergey Tulyakov, Mohit Bansal, Francesco Barbieri, and Yuwei Fang.

, and Dong Yu. 2025. LongMemEval: Benchmarking chat assistants on long-term interactive memory. p. 9
Where the paper says it

Di Wu, Hongwei Wang, Wenhao Yu, Yuwei Zhang, KaiWei Chang, and Dong Yu. LongMemEval: Benchmarking chat assistants on long-term interactive memory.

In The Thirteenth International Conference on Learning Representations. p. 9
Where the paper says it

LongMemEval: Benchmarking chat assistants on long-term interactive memory. In The Thirteenth International Conference on Learning Representations.

Zhiheng Xi, Wenxiang Chen, Xin Guo, Wei He, Yiwen Ding, Boyang Hong, Ming Zhang, Junzhe Wang. p. 9

Where the paper says it

In The Thirteenth International Conference on Learning Representations. Zhiheng Xi, Wenxiang Chen, Xin Guo, Wei He, Yiwen Ding, Boyang Hong, Ming Zhang, Junzhe Wang, Senjie Jin, and 1 others.

, Senjie Jin. p. 9
Where the paper says it

In The Thirteenth International Conference on Learning Representations. Zhiheng Xi, Wenxiang Chen, Xin Guo, Wei He, Yiwen Ding, Boyang Hong, Ming Zhang, Junzhe Wang, Senjie Jin, and 1 others.

, and 1 others. 2023. The rise and potential of large language model based agents: A survey. p. 9
Where the paper says it

Zhiheng Xi, Wenxiang Chen, Xin Guo, Wei He, Yiwen Ding, Boyang Hong, Ming Zhang, Junzhe Wang, Senjie Jin, and 1 others. The rise and potential of large language model based agents: A survey.

Preprint.

Guangxuan Xiao, Yuandong Tian, Beidi Chen, Song Han, and Mike Lewis. p. 9

Where the paper says it

Preprint, arXiv:2309. Guangxuan Xiao, Yuandong Tian, Beidi Chen, Song Han, and Mike Lewis.

2024. Efficient streaming language models with attention sinks. p. 9
Where the paper says it

Guangxuan Xiao, Yuandong Tian, Beidi Chen, Song Han, and Mike Lewis. Efficient streaming language models with attention sinks.

In The Twelfth International Conference on Learning Representations.

Wujiang Xu, Zujie Liang, Kai Mei, Hang Gao, Juntao Tan. p. 9

Where the paper says it

In The Twelfth International Conference on Learning Representations. Wujiang Xu, Zujie Liang, Kai Mei, Hang Gao, Juntao Tan, and Yongfeng Zhang. p. 9

Where the paper says it

In The Twelfth International Conference on Learning Representations. Wujiang Xu, Zujie Liang, Kai Mei, Hang Gao, Juntao Tan, and Yongfeng Zhang.

, and Yongfeng Zhang. 2025. A-MEM: Agentic memory for LLM agents. p. 9
Where the paper says it

Wujiang Xu, Zujie Liang, Kai Mei, Hang Gao, Juntao Tan, and Yongfeng Zhang. A-MEM: Agentic memory for LLM agents.

In Advances in Neural Information Processing Systems.

Wanjun Zhong, Lianghong Guo, Qiqi Gao, He Ye, and Yanlin Wang. p. 9

Where the paper says it

In Advances in Neural Information Processing Systems. Wanjun Zhong, Lianghong Guo, Qiqi Gao, He Ye, and Yanlin Wang.

2023. MemoryBank: Enhancing large language models with long-term memory. p. 9
Where the paper says it

Wanjun Zhong, Lianghong Guo, Qiqi Gao, He Ye, and Yanlin Wang. MemoryBank: Enhancing large language models with long-term memory.

Preprint.

Zijian Zhou, Ao Qu, Zhaoxuan Wu, Sunghwan Kim, Alok Prakash, Daniela Rus, Jinhua Zhao, Bryan Kian Hsiang Low. p. 9

Where the paper says it

Preprint, arXiv:2305. Zijian Zhou, Ao Qu, Zhaoxuan Wu, Sunghwan Kim, Alok Prakash, Daniela Rus, Jinhua Zhao, Bryan Kian Hsiang Low, and Paul Pu Liang. p. 9

Where the paper says it

Preprint, arXiv:2305. Zijian Zhou, Ao Qu, Zhaoxuan Wu, Sunghwan Kim, Alok Prakash, Daniela Rus, Jinhua Zhao, Bryan Kian Hsiang Low, and Paul Pu Liang.

, and Paul Pu Liang. 2025. MEM1: Learning to synergize memory and reasoning for efficient long-horizon agents. p. 9
Where the paper says it

Zijian Zhou, Ao Qu, Zhaoxuan Wu, Sunghwan Kim, Alok Prakash, Daniela Rus, Jinhua Zhao, Bryan Kian Hsiang Low, and Paul Pu Liang. MEM1: Learning to synergize memory and reasoning for efficient long-horizon agents.

Preprint.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Atomic Fact Extraction Module Improvement: Enhance coreference resolution and temporal anchoring in the Fact Executor to more reliably produce self-contained atomic facts, ensuring that each generated fact is independent and comprehensible without external context.

  2. Memory Graph Association Refinement: Improve the Entity Edge weight calculation using a more sophisticated function like wkw(Fi, Fj) = IDF(k) · βq(k) · π(k) to better balance semantic density and keyword overlap while aggressively penalizing frequent noninformative terms.

  3. Episodic Consolidation Enhancement: Refine the Event Memory Construction algorithm by making the LLM's routing decisions more robust, allowing it to better select existing events to absorb Fnew or select standalone facts to trigger the creation of new events when Fnew provides diverse episodic contexts.

  4. State Evolution Stability: Strengthen Temporal Profile Modeling by improving the LLM-driven updater logic during batch processing, ensuring that when a profile changes, AtomMem correctly handles copying previous states into history H and updating historical versions to preserve past user states.

  5. Hierarchical Retrieval Precision: Optimize the Compensatory Recall stage by tuning the compensatory fusion weight we to better balance global event relevance and local fact precision, aiming for an optimal peak at we = 0.7 across all core evaluation dimensions.

Abstract

Large language models (LLMs) demonstrate strong reasoning and generation abilities, but their fixed context windows limit long-term information accumulation and reuse across multi-session interactions. Existing memory-augmented systems often construct memory in a coarse and unstable manner, relying on inefficient memory representations or unstable unconstrained updates. To address these challenges, we propose AtomMem, a long-term memory system designed for value-dense storage and stable memory evolution. AtomMem introduces a Fact Executor, which selectively extracts high value atomic facts from long form interactions to serve as highly efficient memory representations. Subsequently, AtomMem organizes these facts into hierarchical event structures and temporal profiles, capturing coherent episodic contexts and tracking dynamically evolving user attributes over time. During retrieval, the system activates an associative memory graph to connect fragmented memories. Experiments on the LoCoMo benchmark confirm that AtomMem achieves state-of-the-art performance across various reasoning tasks, offering a scalable and economically viable solution for deploying intelligent personalized agents.

Sources

Related papers