EMBERS: Empirically Motivated Robustness Testing of Deep Research Systems via Defect-Guided Fuzzing
cs.AI, cs.LG
Submitted: 2026-11-01
Updated: 2026-11-01
License: http://creativecommons.org/licenses/by/4.0/
The gist: Deep Research (DR) systems extend large language models into autonomous agents that decompose complex information needs, iteratively search the web…
Terminology
Abstract
Deep Research (DR) systems extend large language models into autonomous agents that decompose complex information needs, iteratively search the web, and synthesize evidence to answer challenging questions. However, their multi-stage execution may introduce robustness risks, as variations in user questions or intermediate search inputs may propagate across research stages and lead to incorrect answers. Existing evaluations focus on task-solving capabilities or robustness against external evidence disturbances, leaving intrinsic weaknesses in DR execution insufficiently explored. Thus, we conduct an empirical study of 1,932 execution trajectories from three representative DR systems and identify nine recurring defect categories across subquestion formulation, online search, and final reasoning. Guided by these findings, we propose EMBERS, an empirically motivated, defect-guided fuzzing framework for robustness testing of DR systems. Starting from correctly answered seed questions, EMBERS generates stage-targeted test inputs using defect-guided mutation, web-search-assisted fidelity assurance, and diversity-aware seed maintenance. Experiments across 16 framework--model--dataset settings show that EMBERS achieves 3.21 times the failure-discovery throughput of the strongest baseline under identical testing budgets. Human and LLM-assisted evaluations demonstrate that most generated inputs preserve the original information need and necessary contextual constraints. Moreover, model fine-tuning experiments show consistent-answer performance improvement, highlighting their value for robustness testing and model improvement.
Sources
- ReST meets ReAct: Self-Improvement for Multi-Step Reasoning LLM Agent
- Evaluating Deep-Search Agents under Hierarchical Web Evidence Poisoning
- On the Worst Prompt Performance of Large Language Models
- MindSearch: Mimicking Human Minds Elicits Deep AI Searcher
- DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
- DeepResearch Bench: A Comprehensive Benchmark for Deep Research Agents
- OpenSeeker: Democratizing Frontier Search Agents by Fully Open-Sourcing Training Data
- Deep Research Bench: Evaluating AI Web Research Agents
- Deep Research Agents: A Systematic Examination And Roadmap
- DeepWideSearch: Benchmarking Depth and Width in Agentic Information Seeking
- WebSailor: Navigating Super-human Reasoning for Web Agent
- WebThinker: Empowering Large Reasoning Models with Deep Research Capability
- WebExplorer: Explore and Evolve for Training Long-Horizon Web Agents
- Testing Question Answering Software with Context-Driven Question Generation
- WebGPT: Browser-assisted question-answering with human feedback
- DRNOISE: Benchmarking Deep Research Agents in Misleading Evidence Environments
- FORGE: Research-Trajectory Hijacking Attacks on Deep Research Agents
- ContextPilot: Teaching Agents for Proactive Context Management via Fine-grained RL
- Agent-Fence: Mapping Security Vulnerabilities Across Deep Research Agents
- DeepDiver: Adaptive Search Intensity Scaling via Open-Web Reinforcement Learning
Related papers
- Multi-Objective Structured Pruning of LLMs for Latency and Model Size Optimization
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Embeddings for Preferences, Not Semantics
- Can We Trust Item Response Theory for AI Evaluation?
- Diagnosing Task Insensitivity in Language Agents