MADA-RL: Multi-Agent Debate-Aware Reinforcement Learning for Parameter-Efficient Reasoning in Compact Models
Large language models achieve strong reasoning performance, but often at prohibitive training cost - a challenge that is especially acute for compact models ($\leq 4 \, \mathrm{B}$ parameters) trained under limited budgets. We introduce MADA-RL, a post-training framework that specializes compact models into generator and critic roles and trains them with a debate-aware learning signal, fine-tuning only a small subset of parameters via LoRA adapters. Our central contribution is a counterfactual critic advantage: a dynamic, role-conditioned baseline that redefines the critic's advantage as its r
Lineage graph
Paper → model → repo connections mined from source citations (Tier-1 exact match).
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- FuzzySimilar title/name (fuzzy) · 59%AgentCore-8B →
“Fuzzy title match (0.73): “MADA-RL: Multi-Agent Debate-Aware Reinforcement Learning for” ≈ “AgentCore-8B””
- FuzzySimilar title/name (fuzzy) · 87%SWE-agent/SWE-agent →
“Fuzzy title match (0.94): “MADA-RL: Multi-Agent Debate-Aware Reinforcement Learning for” ≈ “SWE-agent/SWE-agent””
- FuzzySimilar title/name (fuzzy) · 87%zhayujie/CowAgent →
“Fuzzy title match (0.94): “MADA-RL: Multi-Agent Debate-Aware Reinforcement Learning for” ≈ “zhayujie/CowAgent””
- FuzzySimilar title/name (fuzzy) · 66%open-multi-agent/open-multi-agent →
“Fuzzy title match (0.78): “MADA-RL: Multi-Agent Debate-Aware Reinforcement Learning for” ≈ “open-multi-agent/open-multi-agent””
- FuzzySimilar title/name (fuzzy) · 59%NousResearch/hermes-agent →
“Fuzzy title match (0.73): “MADA-RL: Multi-Agent Debate-Aware Reinforcement Learning for” ≈ “NousResearch/hermes-agent””
- FuzzySimilar title/name (fuzzy) · 59%bojieli/ai-agent-book →
“Fuzzy title match (0.73): “MADA-RL: Multi-Agent Debate-Aware Reinforcement Learning for” ≈ “bojieli/ai-agent-book””
- LinkedLinked via arxiv author · 85%Martino M. L. Pulici →
“MADA-RL: Multi-Agent Debate-Aware Reinforcement Learning for Parameter-Efficient Reasoning in Compact Models”
- LinkedLinked via arxiv author · 85%Cuong Xuan Chu →
“MADA-RL: Multi-Agent Debate-Aware Reinforcement Learning for Parameter-Efficient Reasoning in Compact Models”