UCPO modifies GRPO with a uniformity penalty over correct solutions to prevent diversity collapse in RLVR, yielding up to 10% higher Pass@64 on AIME24 and 45% more equation-level diversity.
arXiv preprint arXiv:2509.21128 , year=
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
years
2026 4representative citing papers
CyclicJudge uses round-robin judge-to-scenario assignment to recover the panel-mean score exactly while using the same number of judge calls as single-judge evaluation.
Gaze Attention groups visual embeddings into selectable regions and dynamically restricts attention to task-relevant ones, matching dense baselines with up to 90% fewer visual KV entries via added context tokens.
DoTS decouples SFT and RLVR training then synthesizes their task vectors at inference time to match integrated training results at ~3% compute cost.
citing papers explorer
-
Uniform-Correct Policy Optimization: Breaking RLVR's Indifference to Diversity
UCPO modifies GRPO with a uniformity penalty over correct solutions to prevent diversity collapse in RLVR, yielding up to 10% higher Pass@64 on AIME24 and 45% more equation-level diversity.
-
CyclicJudge: Mitigating Judge Bias Efficiently in LLM-based Evaluation
CyclicJudge uses round-robin judge-to-scenario assignment to recover the panel-mean score exactly while using the same number of judge calls as single-judge evaluation.
-
Learning to See What You Need: Gaze Attention for Multimodal Large Language Models
Gaze Attention groups visual embeddings into selectable regions and dynamically restricts attention to task-relevant ones, matching dense baselines with up to 90% fewer visual KV entries via added context tokens.
-
Decouple before Integration: Test-time Synthesis of SFT and RLVR Task Vectors
DoTS decouples SFT and RLVR training then synthesizes their task vectors at inference time to match integrated training results at ~3% compute cost.