When Should an Agent Refresh? Adaptive Index Reuse for Efficient Long-Horizon Reasoning
cs.LG, cs.CL
Submitted: 2026-11-01
Updated: 2026-11-01
Code: https://github.com/LV-NUS/SFI
Terminology
Sources
- Program Synthesis with Large Language Models
- IndexCache: Accelerating Sparse Attention via Cross-Layer Index Reuse
- $\tau^2$-Bench: Evaluating Conversational Agents in a Dual-Control Environment
- Reducing Transformer Key-Value Cache Size with Cross-Layer Attention
- The MiniMax-M2 Series: Mini Activations Unleashing Max Real-World Intelligence
- DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
- Kascade: A Practical Sparse Attention Method for Long-Context LLM Inference
- HySparse: A Hybrid Sparse Attention Architecture with Oracle Token Selection and KV Cache Sharing
- Gemma 3 Technical Report
- Language Models Can Control Their Own Attention
- MiniMax Sparse Attention
- Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces
- Mixture-of-Depths: Dynamically allocating compute in transformer-based language models
- You Only Cache Once: Decoder-Decoder Architectures for Language Models
- You Only Index Once: Cross-Layer Sparse Attention with Shared Routing
- Reward Constrained Policy Optimization
- LatentIndex: Cross-Layer Sharing with Layer-Specific Selection for Sparse Attention
- HySparse2: Hybrid Sparse Attention with Two-Level KV Sharing
- Slow-Fast Inference: Training-Free Inference Acceleration via Within-Sentence Support Stability
- HISA: Efficient Hierarchical Indexing for Fine-Grained Sparse Attention
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis