Paper ties RL memory retrieval and credit to value of information
A learned world model estimates that quantity from archived rollouts, defining an optimal retrieval set and a TD-target floor.
Stories per day · 30 days
1 RSS
Relatedreinforcement-learning6llm-agents2qwen2agentic-rl1agent-training1browsecomp1diffusion-language-models1gdpo1hotpotqa1language-agents1learning-theory1llada1locomo1math-reasoning1memory1mrtpo1multi-agent1multi-reward1
21 stories, newest first · one card per story
Stories tagged #credit-assignment
Intermittent noisy gradients can leave learning variance unbounded
Average dynamics may still reach the correct optimum while learning variance grows without bound.
FOCUS uses selective branching to train long-horizon agent critics
At Qwen3.5-9B, held-out accuracy rises 7.2 points on BrowseComp-Plus and 4.1 on DeonticBench versus the strongest baseline.
MPO aligns LLM agent credit assignment at the turn level
The method assigns credit at assistant-turn granularity while keeping finer token-level allocation, outperforming RL baselines on agentic tasks.
DTCA assigns token-level credit for RL on diffusion language models
On LLaDA-8B-Instruct, authors report gains of up to 9.0 points on MATH500 and 5.3 on GSM8K.
Self-Credit densifies sparse rewards for long-horizon agentic RL
It raises Qwen3 8B and 32B by 37.0 and 36.2 points on BrowseComp-Plus over the base models.
GRSE credits multi-agent skills by contrasting repeated runs
On HotpotQA and LoCoMo it beat MASkills on most categories, with the largest gains on multi-hop questions.
Each Reward to Its Relevant Tokens: Reward-Specific Token-Level Credit Assignment for Multi-Reward RL
MRTPO assigns each reward's advantage only to tokens relevant to that verifier, improving multi-reward RL.
Beyond Temporal Proximity: Explanatory World Models for Credit Assignment in Reinforcement Learning
Explanatory World Models reassign delayed RL credit and beat PPO and LaRe on long-horizon tasks.
Pivot-SD: Efficient Self-Distillation for Masked Diffusion Language Models
Pivot-SD trains masked diffusion LLMs only on high-impact token commitments, improving LLaDA-8B-Instruct math and code results with just 200 questions.
Credit Where It Matters: Dependency-Aware Policy Optimization for Terminal Agents
DepGPO improves RL for terminal agents by building command dependency graphs from execution traces to assign credit to reads and writes that actually affect task outcomes.
Fixing GRPO's credit assignment problem without evaluating every step
An arXiv paper claims to fix GRPO credit assignment without scoring every training step.
Semifactual Credit-Augmented Policy Optimization
SCAPO uses semifactual token stability to beat GRPO on AIME for two Qwen3 base models.
Semifactual Credit-Augmented Policy Optimization
SCAPO adds semifactual token stability to GRPO, lifting Qwen3 AIME accuracy by up to 5.63 points.
Can We Trust the Teacher? Decoupled Credit Direction-Magnitude for Self-Distillation
A new self-distillation method separates credit direction from magnitude and beats prior RL baselines.
Surprising Success, Repeated Failure: Entropy-Guided Credit Assignment for Exploration in LLM Reasoning
EAPO uses policy entropy to reinforce surprising successes and penalize repeated failures in LLM reasoning.
Groupwise Agentic Grading and Advantage Redistribution for Code Agent RL
GAGAR uses groupwise agentic grading to shift code-agent RL credit toward cleaner implementations.
HySTAR: Anchored Hypergraphs for Stable Credit Assignment in Cooperative Multi-Agent Reinforcement Learning
HySTAR anchors a fixed hypergraph so cooperative MARL assigns credit without structural target drift.
SLCA-GRPO: Resolving Cross-Segment Credit Misattribution in Tool-Calling RL
SLCA-GRPO assigns separate advantages to tool and summary tokens, improving tool-calling reinforcement learning.
When and Where to Trust the Teacher: Unifying On-Policy Distillation and GRPO through Entropy-Calibrated Credit Assignment
UECR-GRPO blends teacher and verifier credit, slightly improving Qwen3 math-reasoning accuracy.
PACT: From Credit Assignment to Critic Alignment
PACT realigns actor-critic training and beats GRPO and PPO on math reasoning and SWE-bench Verified.