VideoP2R separates perception and reasoning in a process-aware RFT pipeline with a new CoT dataset and PA-GRPO rewards, reaching SOTA on six of seven video benchmarks.
Edge-grpo: Entropy-driven grpo with guided error correction for advantage diversity.arXiv preprint arXiv:2507.21848
11 Pith papers cite this work. Polarity classification is still indexing.
11
Pith papers citing it
citation-role summary
background 2
citation-polarity summary
roles
background 2polarities
background 2representative citing papers
Introduces RSI metric and RSI-S filtering method for adaptive token selection in RLVR, reporting 2-3 point gains over GRPO on AIME/AMC benchmarks.
Reinforcement learning after SFT conversion narrows the performance gap between sliding-window attention and full self-attention on math reasoning benchmarks while preserving linear complexity.
Smaller models provide temporally correlated policy-level diversity that serves as structured exploration for training larger models in GRPO, yielding accuracy gains such as +8.8% on AIME 24 with reduced compute via the S2L-PO framework.
GRPO suffers advantage collapse on uniform-reward groups; ACR quantifies it and AVSPO adds virtual samples to restore gradients, yielding 4-6% accuracy gains on math benchmarks across 0.5B-14B models.
Span-level Wasserstein distances between hidden-state distributions of correct and incorrect rollouts provide a self-supervised signal to reweight advantages in GRPO, improving fine-grained credit assignment on math and code tasks.
SCOPE-RL adds a regularization term built from high-temperature positive samples to quantitatively control entropy dynamics and maintain exploration in RL post-training of reasoning LLMs.
Survey that defines agentic RL for LLMs via POMDPs, introduces a taxonomy of planning/tool-use/memory/reasoning capabilities and domains, and compiles open environments from over 500 papers.
EDAS modulates RL advantage signals for incorrect rollouts by amplifying penalties on repeated errors and attenuating them on rare ones, yielding average gains of 6.29 points over DAPO on Qwen3-8B across seven math benchmarks.
StaRPO improves LLM reasoning by adding autocorrelation function and path efficiency stability metrics to RL policy optimization, yielding higher accuracy and fewer logic errors on reasoning benchmarks.
ConsistRoll enforces cross-view consistency during RLVR training for MLLMs by joint rewards on grouped original and augmented views, yielding robustness gains on math, general, and hallucination benchmarks.
citing papers explorer
-
VIDEOP2R: Video Understanding from Perception to Reasoning
VideoP2R separates perception and reasoning in a process-aware RFT pipeline with a new CoT dataset and PA-GRPO rewards, reaching SOTA on six of seven video benchmarks.
-
Which Tokens Matter? Adaptive Token Selection for RLVR with the Relative Surprisal Index
Introduces RSI metric and RSI-S filtering method for adaptive token selection in RLVR, reporting 2-3 point gains over GRPO on AIME/AMC benchmarks.
-
Architecture-Aware Reinforcement Learning Makes Sliding-Window Attention Competitive in Math Reasoning
Reinforcement learning after SFT conversion narrows the performance gap between sliding-window attention and full self-attention on math reasoning benchmarks while preserving linear complexity.
-
Smaller Models are Natural Explorers for Policy-Level Diversity in GRPO
Smaller models provide temporally correlated policy-level diversity that serves as structured exploration for training larger models in GRPO, yielding accuracy gains such as +8.8% on AIME 24 with reduced compute via the S2L-PO framework.
-
Advantage Collapse in Group Relative Policy Optimization: Diagnosis and Mitigation
GRPO suffers advantage collapse on uniform-reward groups; ACR quantifies it and AVSPO adds virtual samples to restore gradients, yielding 4-6% accuracy gains on math benchmarks across 0.5B-14B models.
-
Hidden States Know Where Reasoning Diverges: Credit Assignment via Span-Level Wasserstein Distance
Span-level Wasserstein distances between hidden-state distributions of correct and incorrect rollouts provide a self-supervised signal to reweight advantages in GRPO, improving fine-grained credit assignment on math and code tasks.
-
SCOPE-RL: Stable and Quantitative Control of Policy Entropy in RL Post-Training
SCOPE-RL adds a regularization term built from high-temperature positive samples to quantitatively control entropy dynamics and maintain exploration in RL post-training of reasoning LLMs.
-
The Landscape of Agentic Reinforcement Learning for LLMs: A Survey
Survey that defines agentic RL for LLMs via POMDPs, introduces a taxonomy of planning/tool-use/memory/reasoning capabilities and domains, and compiles open environments from over 500 papers.
-
Leveraging Error Diversity in Group Rollouts for Reinforcement Learning
EDAS modulates RL advantage signals for incorrect rollouts by amplifying penalties on repeated errors and attenuating them on rare ones, yielding average gains of 6.29 points over DAPO on Qwen3-8B across seven math benchmarks.
-
StaRPO: Stability-Augmented Reinforcement Policy Optimization
StaRPO improves LLM reasoning by adding autocorrelation function and path efficiency stability metrics to RL policy optimization, yielding higher accuracy and fewer logic errors on reasoning benchmarks.
-
Consistency as Inductive Bias: Learning Cross-View Invariance for Robust Multimodal Reasoning
ConsistRoll enforces cross-view consistency during RLVR training for MLLMs by joint rewards on grouped original and augmented views, yielding robustness gains on math, general, and hallucination benchmarks.