mlsanityData to 7 Oct 2026

TRACE: Capability-Targeted Agentic Training

arXiv 2604.05336 ↗First posted 7 Apr 2026cs.AIcited by 2

On the map: LLM agents and tool use → General LLM agent frameworks and benchmarks

Abstract (arXiv)

Models often fail to complete agentic tasks because they lack core capabilities required by the target environment. However, mainstream approaches for addressing these failures typically either fine-tune directly on target environments or generate synthetic data that is not targeted to the model's actual capability deficits, resulting in low sample efficiency and limited generalization. We introduce TRACE (Turning Recurrent Agent failures into Capability-targeted training Environments), an end-to-end system for environment-specific agent self-improvement. TRACE contrasts successful and failed trajectories to automatically identify missing capabilities, synthesizes a targeted training environment for each capability that rewards whether the capability is exercised, trains a LoRA adapter via reinforcement learning on each synthetic environment, and then trains a mixture-of-experts model over the capability adapters. TRACE can be effectively applied across different environments, improving over the base agent by +15.3 points on τ2-Bench, a customer-service agent benchmark, and by +15.0 points Pass@1 on SWE-Bench Verified, a software-engineering benchmark. TRACE outperforms the strongest external baselines, GEPA and SWE-RL, by +8.6 points and +8.4 points, respectively. In addition, TRACE is more sample-efficient than strong fine-tuning baselines: using fewer than one-fourth the number of rollouts, TRACE outperforms the best-performing baselines, GRPO and GEPA, and achieves higher final accuracy by +10.4 and +8.6 points on τ2-Bench.

Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Placed in General LLM agent frameworks and benchmarks, one of 16 directions in LLM agents and tool use. Builds on 28 of its 40 references that are in the map. First posted on arXiv 7 Apr 2026.

Builds on

28 of 40 references.

Closest ancestors · 5

Same direction, newest first

  1. 2026Agent World Model: Infinity Synthetic Environments for Agentic Reinforcement Learning10 Feb
  2. 2025Agent Data Protocol: Unifying Datasets for Diverse, Effective Fine-tuning of LLM Agents28 Oct
  3. 2025Towards General Agentic Intelligence via Environment Scaling16 Sep
  4. 2024TheAgentCompany: Benchmarking LLM Agents on Consequential Real World Tasks18 Dec
  5. 2024τ-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains17 Jun

Other references · 23

  1. 2017Proximal Policy Optimization Algorithmscited by 14,441
  2. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,084
  3. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  4. 2017Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layercited by 3,413
  5. 2023SWE-bench: Can Language Models Resolve Real-World GitHub Issues?cited by 3,193
  6. 2022Editing Models with Task Arithmeticcited by 1,076
  7. 2023TIES-Merging: Resolving Interference When Merging Modelscited by 795
  8. 2025τ2-Bench: Evaluating Conversational Agents in a Dual-Control Environmentcited by 558
  9. 2024WorkArena: How Capable Are Web Agents at Solving Common Knowledge Work Tasks?cited by 348
  10. 2024ToolSandbox: A Stateful, Conversational, Interactive Evaluation Benchmark for LLM Tool Use Capabilitiescited by 272
  11. 2024DigiRL: Training In-The-Wild Device-Control Agents with Autonomous Reinforcement Learningcited by 210
  12. 2023Pushing Mixture of Experts to the Limit: Extremely Parameter Efficient MoE for Instruction Tuningcited by 148
11 more
  1. 2024Higher Layers Need More LoRA Expertscited by 101
  2. 2025AgentRL: Scaling Agentic Reinforcement Learning with a Multi-Turn, Multi-Task Frameworkcited by 58
  3. 2024MoELoRA: Contrastive Learning Guided Mixture of Experts on Parameter-Efficient Fine-Tuning for Large Language Modelscited by 54
  4. 2026EnvScaler: Scaling Tool-Interactive Environments for LLM Agent via Programmatic Synthesiscited by 46
  5. 2025Agentic Reinforcement Learning with Implicit Step Rewardscited by 45
  6. 2025Accurate and Efficient Low-Rank Model Merging in Core Spacecited by 28
  7. 2026ScaleEnv: Scaling Environment Synthesis from Scratch for Generalist Interactive Tool-Use Agent Trainingcited by 23
  8. 2025Procedural Environment Generation for Tool-Use Agentscited by 21
  9. 2026From Self-Evolving Synthetic Data to Verifiable-Reward RL: Post-Training Multi-turn Interactive Tool-Using Agentscited by 17
  10. 2025Each Rank Could be an Expert: Single-Ranked Mixture of Experts LoRA for Multi-Task Learningcited by 12
  11. 2026CoMoL: Efficient Mixture of LoRA Experts via Dynamic Core Space Mergingcited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026Terminal Agents: A Survey of AI Agents in Command-Line Environmentscited by 0
  2. 2026Behavior Quotient Learning for Low-Rank Adaptation of LLM Agentscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026TraceDance: An Automated System for Building Agent Behavior Benchmarks from Real-World Agent Deployment Traces
  2. 2026SkillFocus: Evolving Agent Skills via Capability Decomposition
  3. 2026TRACE: A Self-Evolving Skill Bank for Consistent, Limit-Aware LLM Agents
  4. 2026EvoAgentBench: Benchmarking Agent Self-Evolution via Ability Transfer
  5. 2026BenchTrace: A Benchmark for Testing Reflection Ability and Controlled Evolution in LLM Agents
  6. 2026SkillComposer: Learning to Evolve Agent Skills for Specification and Generalization
  7. 2026Agent Plasticity: Measuring Self-Improvement Through Experience
  8. 2026Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills