Exact Is Easier: Credit Assignment for Cooperative LLM Agents

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Chen, Yanjun, Sun, Yirong, Wang, Hanlin, Wang, Jinghan, Zhang, Xinming, Shen, Xiaoyu, Li, Wenjie, Zhang, Wei
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866911662410301440
author Chen, Yanjun
Sun, Yirong
Wang, Hanlin
Wang, Jinghan
Zhang, Xinming
Shen, Xiaoyu
Li, Wenjie
Zhang, Wei
author_facet Chen, Yanjun
Sun, Yirong
Wang, Hanlin
Wang, Jinghan
Zhang, Xinming
Shen, Xiaoyu
Li, Wenjie
Zhang, Wei
contents Removing an agent from a cooperative team to measure its contribution seems natural, yet in multi-agent LLM systems this evaluation distorts the result it claims to measure. This failure is not isolated: learned critics, trajectory-level baselines, and agent-removal counterfactuals all inherit from standard multi-agent reinforcement learning a premise that exact counterfactual evaluation requires privileged environment access, and therefore approximate. In cooperative LLM systems, this premise is false. Interaction histories are deterministic functions of observable text with no hidden state, so any decision point can be restored exactly, making direct causal measurement possible without parametric approximation. C3 exploits this property by fixing the complete history at each decision point, sampling alternative actions under a frozen behavior policy, and computing unbiased per-decision advantages through a parameter-free leave-one-out baseline. Across six benchmarks spanning math reasoning and code generation, two model families, and two multi-agent topologies, C3 consistently outperforms all baselines; a controlled decomposition confirms gains originate from credit quality, not architecture, while checkpoint restoration reduces training token consumption. The exact solution proves simpler, cheaper, and more effective than all approximate alternatives. The same structural property that enables exact credit also enables exact verification: three independently computable diagnostics, credit fidelity, within-group variance, and inter-agent influence, constitute the first method-agnostic auditing tool for multi-agent LLM credit assignment. Our code is available at https://github.com/EIT-EAST-Lab/C3
format Preprint
id arxiv_https___arxiv_org_abs_2603_06859
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Exact Is Easier: Credit Assignment for Cooperative LLM Agents
Chen, Yanjun
Sun, Yirong
Wang, Hanlin
Wang, Jinghan
Zhang, Xinming
Shen, Xiaoyu
Li, Wenjie
Zhang, Wei
Machine Learning
Artificial Intelligence
Removing an agent from a cooperative team to measure its contribution seems natural, yet in multi-agent LLM systems this evaluation distorts the result it claims to measure. This failure is not isolated: learned critics, trajectory-level baselines, and agent-removal counterfactuals all inherit from standard multi-agent reinforcement learning a premise that exact counterfactual evaluation requires privileged environment access, and therefore approximate. In cooperative LLM systems, this premise is false. Interaction histories are deterministic functions of observable text with no hidden state, so any decision point can be restored exactly, making direct causal measurement possible without parametric approximation. C3 exploits this property by fixing the complete history at each decision point, sampling alternative actions under a frozen behavior policy, and computing unbiased per-decision advantages through a parameter-free leave-one-out baseline. Across six benchmarks spanning math reasoning and code generation, two model families, and two multi-agent topologies, C3 consistently outperforms all baselines; a controlled decomposition confirms gains originate from credit quality, not architecture, while checkpoint restoration reduces training token consumption. The exact solution proves simpler, cheaper, and more effective than all approximate alternatives. The same structural property that enables exact credit also enables exact verification: three independently computable diagnostics, credit fidelity, within-group variance, and inter-agent influence, constitute the first method-agnostic auditing tool for multi-agent LLM credit assignment. Our code is available at https://github.com/EIT-EAST-Lab/C3
title Exact Is Easier: Credit Assignment for Cooperative LLM Agents
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2603.06859