mlsanityData to 7 Oct 2026

TRACE: Trajectory Reasoning through Adaptive Cross-Step Evidence Aggregation for LLM Agents

arXiv 2606.07054 ↗First posted 5 Jun 2026cs.CL cs.AI cs.CRcited by 3

On the map: LLM reasoning → Reasoning for agents, planning and code

also in 1 more
Abstract (arXiv)

Autonomous LLM agents can pursue hidden malicious objectives through sequences of individually benign actions, making sabotage difficult to detect using standard trajectory-level monitoring. Existing approaches either evaluate complete trajectories in a single pass or partition them into independently scored windows, limiting their ability to connect evidence across temporally distant actions. We propose TRACE, a monitoring framework for long-horizon LLM agent trajectories. TRACE operates through a TIJ (Triage-Inspect-Judge) loop that identifies high-signal regions, performs targeted inspection while maintaining accumulated evidence across reasoning steps, and synthesizes a trajectory-level verdict. We evaluate TRACE on ten task domains from SHADE-Arena against state-of-the-art baselines. TRACE achieves an aggregate F1 of 0.713 and recall of 0.844, with the largest gains on tasks requiring long-range evidence linking.

Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Placed in Reasoning for agents, planning and code, one of 17 directions in LLM reasoning. Builds on 30 of its 41 references that are in the map. First posted on arXiv 5 Jun 2026.

Builds on

30 of 41 references.

Closest ancestors · 1

Same direction, newest first

  1. 2025AgentAuditor: Human-Level Safety and Security Evaluation for LLM Agents31 May

Other references · 29

  1. 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,355
  2. 2023Reflexion: Language Agents with Verbal Reinforcement Learningcited by 4,064
  3. 2023Generative Agents: Interactive Simulacra of Human Behaviorcited by 3,500
  4. 2023Universal and Transferable Adversarial Attacks on Aligned Language Modelscited by 2,914
  5. 2023Voyager: An Open-Ended Embodied Agent with Large Language Modelscited by 2,081
  6. 2023AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversationcited by 1,906
  7. 2023CAMEL: Communicative Agents for "Mind" Exploration of Large Language Model Societycited by 1,348
  8. 2023Measuring Faithfulness in Chain-of-Thought Reasoningcited by 668
  9. 2023Identifying the Risks of LM Agents with an LM-Emulated Sandboxcited by 514
  10. 2024Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Trainingcited by 478
  11. 2023Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervisioncited by 445
  12. 2024AgentHarm: A Benchmark for Measuring Harmfulness of LLM Agentscited by 406
17 more
  1. 2023Formalizing and Benchmarking Prompt Injection Attacks and Defensescited by 328
  2. 2024Alignment faking in large language modelscited by 316
  3. 2024TheAgentCompany: Benchmarking LLM Agents on Consequential Real World Taskscited by 306
  4. 2024Agent-SafetyBench: Evaluating the Safety of LLM Agentscited by 280
  5. 2024Frontier Models are Capable of In-context Schemingcited by 252
  6. 2024Chain-of-Thought Reasoning Without Promptingcited by 208
  7. 2023AI Control: Improving Safety Despite Intentional Subversioncited by 207
  8. 2024AgentDojo: A Dynamic Environment to Evaluate Prompt Injection Attacks and Defenses for LLM Agentscited by 129
  9. 2023Scheming AIs: Will AIs fake alignment during training in order to get power?cited by 90
  10. 2025AGrail: A Lifelong Agent Guardrail with Effective and Adaptive Safety Detectioncited by 82
  11. 2025Auditing language models for hidden objectivescited by 68
  12. 2024Towards evaluations-based safety cases for AI schemingcited by 32
  13. 2026Compound Deception in Elite Peer Review: A Failure Mode Taxonomy of 100 Fabricated Citations at NeurIPS 2025cited by 13
  14. 2026How does information access affect LLM monitors' ability to detect sabotage?cited by 12
  15. 2025Sentinel Agents for Secure and Trustworthy Agentic AI in Multi-Agent Systemscited by 7
  16. 2026Constitutional Black-Box Monitoring for Scheming in LLM Agentscited by 7
  17. 2025Towards Machine Theory of Mind with Large Language Model-Augmented Inverse Planningcited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026What Resolve Rate Hides: Trajectory Structure Diagnostics for Coding Agentscited by 5
  2. 2026Doomed from the Start: Early Abort of LLM Agent Episodes via a Recall-Controlled Probe Cascadecited by 4
  3. 2026On the Effectiveness of Kernel-Level Evidence for Agent Securitycited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026TRACE: Trajectory Risk-Aware Compression for Long-Horizon Agent Safety
  2. 2026TRACES: Proactive Safety Auditing for Multi-Turn LLM Agents via Trajectory-State Modeling
  3. 2025Beyond the Final Answer: Evaluating the Reasoning Trajectories of Tool-Augmented Agents
  4. 2026SkillTV-Bench: Benchmarking How Well Judges Perform on Skill-Augmented Agentic Execution
  5. 2026Agent-ToM: Learning to Monitor Autonomous LLM Agents via Theory-of-Mind Reasoning
  6. 2026FALAT: Tracing Failures in LLM Agent Trajectories via Dependency-Guided Search
  7. 2026Trust No Tool: Evaluating and Defending LLM Agents under Untrusted Tool Feedback
  8. 2026Long-Horizon Agent Trajectory Attribution: A Unified Benchmark and Fine-Grained Annotation Framework