Of Errors and Echoes: Responsibility-Aware Credit Assignment for Tool-Use Agents
cs.CL, cs.AI
Submitted: 2026-10-12
Updated: 2026-10-12
Comments: 29 pages
Code: https://github.com/xid32/ECHO
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- RUDDER: Return Decomposition for Delayed Rewards
- $\tau^2$-Bench: Evaluating Conversational Agents in a Dual-Control Environment
- Why Do Multi-Agent LLM Systems Fail?
- ACEBench: Who Wins the Match Point in Tool Usage?
- Responsibility and blame: a structural-model approach
- Agentic Reinforced Policy Optimization
- ReTool: Reinforcement Learning for Strategic Tool Use in LLMs
- DRACO: Fine-Grained Credit Assignment with Dynamic Rubrics for Long-Horizon Agent Training
- CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquing
- Simulating Environments with Reasoning Models for Agent Training
- TCPO: Turn-Level Credit Policy Optimization
- Let's Verify Step by Step
- Agent Lightning: Train ANY AI Agents with Reinforcement Learning
- CRAFT: Counterfactual Credit Assignment from Free Sibling Rollouts for Self-Distilled Agentic Reinforcement Learning
- Would I have gotten that reward? Long-term credit assignment by counterfactual contribution analysis
- ToolRL: Reward is All Tool Learning Needs
- Proximal Policy Optimization Algorithms
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- Hindsight Credit Assignment for Long-Horizon LLM Agents
- Process-Supervised Reinforcement Learning for Interactive Multimodal Tool-Use Agents
Related papers
- Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
- Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
- Subliminal Steering: Stronger Encoding of Hidden Signals
- MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
- The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
- Untangling the Mechanisms of Misleading Context in Medical Question Answering