SDRL trains LLMs via self-generated multi-path debates and joint optimization of standalone plus debate-conditioned responses to boost both single-model reasoning and multi-agent debate performance.
Canonical reference
Llava-critic-r1: Your critic model is secretly a strong policy model.arXiv preprint arXiv:2509.00676, 2025b
Canonical reference. 80% of citing Pith papers cite this work as background.
7
Pith papers citing it
Background
80% of classified citations
citation-role summary
background 4
baseline 1
citation-polarity summary
representative citing papers
Introduces VURB benchmark and VUP-35K dataset to train discriminative and generative video reward models that achieve SOTA performance on VURB and VideoRewardBench.
DeScore decouples CoT reasoning from reward scoring in video reward models using a two-stage training process to improve generalization and avoid optimization bottlenecks of coupled generative RMs.
SSL-R1 reformulates visual SSL tasks into verifiable puzzles to supply rewards for RL post-training of MLLMs, yielding gains on multimodal benchmarks without external supervision.
Filtering post-training data to visually grounded questions improves VLM video understanding performance by up to 6.2 points using 69% of the data.
DT2IT-MRM proposes a debiased preference construction pipeline, T2I data reformulation, and iterative training to curate multimodal preference data, achieving SOTA on VL-RewardBench, Multimodal RewardBench, and MM-RLHF-RewardBench.
citing papers explorer
-
Learning from Self-Debate: Preparing Reasoning Models for Multi-Agent Debate
SDRL trains LLMs via self-generated multi-path debates and joint optimization of standalone plus debate-conditioned responses to boost both single-model reasoning and multi-agent debate performance.
-
Video Understanding Reward Modeling: A Robust Benchmark and Performant Reward Models
Introduces VURB benchmark and VUP-35K dataset to train discriminative and generative video reward models that achieve SOTA performance on VURB and VideoRewardBench.
-
Think, then Score: Decoupled Reasoning and Scoring for Video Reward Modeling
DeScore decouples CoT reasoning from reward scoring in video reward models using a two-stage training process to improve generalization and avoid optimization bottlenecks of coupled generative RMs.
-
SSL-R1: Self-Supervised Visual Reinforcement Post-Training for Multimodal Large Language Models
SSL-R1 reformulates visual SSL tasks into verifiable puzzles to supply rewards for RL post-training of MLLMs, yielding gains on multimodal benchmarks without external supervision.
-
Watch Before You Answer: Learning from Visually Grounded Post-Training
Filtering post-training data to visually grounded questions improves VLM video understanding performance by up to 6.2 points using 69% of the data.
-
DT2IT-MRM: Debiased Preference Construction and Iterative Training for Multimodal Reward Modeling
DT2IT-MRM proposes a debiased preference construction pipeline, T2I data reformulation, and iterative training to curate multimodal preference data, achieving SOTA on VL-RewardBench, Multimodal RewardBench, and MM-RLHF-RewardBench.
- High-Entropy Tokens as Multimodal Failure Points in Vision-Language Models