{"total":22,"items":[{"citing_arxiv_id":"2606.20881","ref_index":14,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"When Do Intrinsic Rewards Work for Code Reasoning? A Comprehensive Study","primary_cat":"cs.AI","submitted_at":"2026-06-18T19:15:50+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Empirical evaluation on LiveCodeBench shows certainty-based RLIF yields early gains followed by output shortening and reasoning collapse, providing no advantage for RLVR initialization on code tasks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.04516","ref_index":36,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"GeoMin: Data-Efficient Semi-Supervised RLVR via Geometric Distribution Modeling","primary_cat":"cs.LG","submitted_at":"2026-06-03T06:47:50+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"GeoMin uses geometric distribution modeling on labeled data to assess self-reward reliability, enabling better performance in semi-supervised RLVR with only 10% of typical annotations.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.01464","ref_index":26,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Cross-lingual Self-Consistency for Multilingual Reasoning with Language Models","primary_cat":"cs.CL","submitted_at":"2026-05-31T21:48:26+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Unsupervised RL enforces cross-lingual self-consistency to improve multilingual math reasoning by up to 21.7% on MGSM without gold answers or parallel data, with generalization to unseen languages.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.01249","ref_index":72,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Trust Region On-Policy Distillation","primary_cat":"cs.LG","submitted_at":"2026-05-31T14:04:51+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"TrOPD stabilizes on-policy distillation for LLMs with trust-region learning, outlier estimation, and off-policy guidance, outperforming prior OPD methods on reasoning and code benchmarks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.25864","ref_index":31,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"When Self-Belief Misleads: Active Label Acquisition for Reinforcement Learning with Verifiable Rewards","primary_cat":"cs.LG","submitted_at":"2026-05-25T13:55:12+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"RLAVR uses the Corrective Advantage Gap metric and CARE policy to actively acquire ground-truth labels for key samples, stabilizing RLVR training and boosting performance with limited annotation budgets.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.21125","ref_index":29,"ref_count":2,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Advantage Collapse in Group Relative Policy Optimization: Diagnosis and Mitigation","primary_cat":"cs.LG","submitted_at":"2026-05-20T12:57:37+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"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.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.17037","ref_index":58,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"D$^2$Evo: Dual Difficulty-Aware Self-Evolution for Data-Efficient Reinforcement Learning","primary_cat":"cs.LG","submitted_at":"2026-05-16T15:16:00+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"D²Evo mines medium-difficulty anchors from the current model, trains a Questioner to generate matching questions, and jointly optimizes Solver and Questioner for progressive gains, outperforming baselines on math reasoning with under 2K real samples.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.11775","ref_index":47,"ref_count":2,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Entropy Polarity in Reinforcement Fine-Tuning: Direction, Asymmetry, and Control","primary_cat":"cs.LG","submitted_at":"2026-05-12T08:47:05+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Entropy polarity is a signed token-level quantity derived from a first-order approximation of entropy change that predicts whether RL updates expand or contract policy entropy in LLM fine-tuning, revealing an asymmetry between high- and low-probability tokens.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.08516","ref_index":14,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"OracleTSC: Oracle-Informed Reward Hurdle and Uncertainty Regularization for Traffic Signal Control","primary_cat":"cs.AI","submitted_at":"2026-05-08T21:55:41+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"OracleTSC introduces a reward hurdle and uncertainty regularization to stabilize LLM-based reinforcement learning for traffic signal control, delivering 75% lower travel time and 67% lower queue length on benchmarks plus cross-intersection generalization.","context_count":1,"top_context_role":"background","top_context_polarity":"unclear","context_text":"(2025) with self-consistency maximization Chen et al. (2025), providing a single, tunable mechanism that encourages both confidence and internal consensus-two properties essential for stabilizing long-horizon control in traffic signal optimization. After incorporating the two mechanisms, our final total reward is then formulated as R(st O,at O) =R env(st O,at O)−HR +wERanswer E ,(14) wherew E is a hyperparameter that balances the impact of uncertainty. 4 Experiments We empirically investigate three key questions:(Q1)Does minimizingTemperature-Scaled Softmax Discrete Semantic Entropyimprove traffic-signal control performance compared to training without an uncertainty regularizer?(Q2)Does introducing a reward hurdle on the environmental reward improve performance by"},{"citing_arxiv_id":"2605.07244","ref_index":69,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Experience Sharing in Mutual Reinforcement Learning for Heterogeneous Language Models","primary_cat":"cs.LG","submitted_at":"2026-05-08T05:01:40+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Mutual Reinforcement Learning allows heterogeneous LLMs to exchange experience through mechanisms like Peer Rollout Pooling, Cross-Policy GRPO Advantage Sharing, and Success-Gated Transfer, with outcome-level sharing identified as favorable on the stability-support trade-off.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.06642","ref_index":43,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"StraTA: Incentivizing Agentic Reinforcement Learning with Strategic Trajectory Abstraction","primary_cat":"cs.CL","submitted_at":"2026-05-07T17:51:16+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"StraTA improves LLM agent success rates to 93.1% on ALFWorld and 84.2% on WebShop by sampling a compact initial strategy and training it jointly with action execution via hierarchical GRPO-style rollouts.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.04542","ref_index":88,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Power Distribution Bridges Sampling, Self-Reward RL, and Self-Distillation","primary_cat":"cs.LG","submitted_at":"2026-05-06T06:42:47+00:00","verdict":"UNVERDICTED","verdict_confidence":"UNKNOWN","novelty_score":6.0,"formal_verification":"none","one_line_summary":"The power distribution is the target of power sampling, the closed-form solution to self-reward KL-regularized RL, and the basis for power self-distillation that matches sampling performance at lower cost.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.01428","ref_index":33,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Hallucinations Undermine Trust; Metacognition is a Way Forward","primary_cat":"cs.CL","submitted_at":"2026-05-02T12:59:14+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"LLMs need metacognition to align expressed uncertainty with their actual knowledge boundaries, moving beyond knowledge expansion to reduce confident errors.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.21327","ref_index":6,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Understanding and Mitigating Spurious Signal Amplification in Test-Time Reinforcement Learning for Math Reasoning","primary_cat":"cs.LG","submitted_at":"2026-04-23T06:32:08+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"DDRL reduces spurious reward noise in test-time RL for math by excluding ambiguous samples, using fixed advantages, and adding consensus-based updates, outperforming prior TTRL methods on math benchmarks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.18493","ref_index":13,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Too Correct to Learn: Reinforcement Learning on Saturated Reasoning Data","primary_cat":"cs.LG","submitted_at":"2026-04-20T16:43:28+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"A parameter-free sampling strategy called CUTS combined with Mixed-CUTS training prevents mode collapse in RL for saturated LLM reasoning tasks and raises AIME25 Pass@1 accuracy by up to 15.1% over standard GRPO.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2605.04065","ref_index":64,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Free Energy-Driven Reinforcement Learning with Adaptive Advantage Shaping for Unsupervised Reasoning in LLMs","primary_cat":"cs.CL","submitted_at":"2026-04-11T07:26:04+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"FREIA applies free energy principles and adaptive advantage shaping to unsupervised RL, outperforming baselines by 0.5-3.5 Pass@1 points on math reasoning with a 1.5B model.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2604.07864","ref_index":30,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"ZeroCoder: Can LLMs Improve Code Generation Without Ground-Truth Supervision?","primary_cat":"cs.SE","submitted_at":"2026-04-09T06:24:54+00:00","verdict":null,"verdict_confidence":null,"novelty_score":null,"formal_verification":null,"one_line_summary":null,"context_count":1,"top_context_role":"background","top_context_polarity":"background","context_text":"Beeching, Agustín Piqueres Lajarín, Quentin Gallouédec, Nathan Habib, Lewis Tunstall, and Leandro von Werra. 2025. CodeForces. https://huggingface.co/ datasets/open-r1/codeforces. [29] Mihir Prabhudesai, Lili Chen, Alex Ippoliti, Katerina Fragkiadaki, Hao Liu, and Deepak Pathak. 2025. Maximizing confidence alone improves reasoning.arXiv preprint arXiv:2505.22660(2025). [30] Archiki Prasad, Elias Stengel-Eskin, Justin Chih-Yao Chen, Zaid Khan, and Mohit Bansal. 2025. Learning to generate unit tests for automated debugging.arXiv preprint arXiv:2502.01619(2025). [31] Olivier Roy and Martin Vetterli. 2007. The effective rank: A measure of effective dimensionality. In2007 15th European signal processing conference. IEEE, 606-610."},{"citing_arxiv_id":"2603.19880","ref_index":11,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"What If Consensus Lies? Selective-Complementary Reinforcement Learning at Test Time","primary_cat":"cs.LG","submitted_at":"2026-03-20T11:47:12+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"SCRL adds selective positive pseudo-labeling and entropy-gated negative pseudo-labeling to test-time RL, reducing noise from weak consensus and improving LLM reasoning on benchmarks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2602.12579","ref_index":21,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"VI-CuRL: Stabilizing Verifier-Independent RL Reasoning via Confidence-Guided Variance Reduction","primary_cat":"cs.LG","submitted_at":"2026-02-13T03:40:52+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"VI-CuRL stabilizes verifier-independent RL for LLM reasoning via confidence-guided curriculum that reduces action and problem variance, with a claimed proof of asymptotic unbiasedness and empirical gains over baselines.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2510.07962","ref_index":16,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"LightReasoner: Can Small Language Models Teach Large Language Models Reasoning?","primary_cat":"cs.CL","submitted_at":"2025-10-09T08:55:12+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"LightReasoner distills supervision signals from SLM-LLM behavioral divergence to improve LLM reasoning on math benchmarks with up to 28.1% accuracy gains and 90-99% reductions in resources.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2509.14234","ref_index":16,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Compute as Teacher: Turning Inference Compute Into Reference-Free Supervision","primary_cat":"cs.LG","submitted_at":"2025-09-17T17:59:42+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Parallel inference rollouts aggregated into pseudo-references enable reference-free RL supervision that matches expert-annotated performance on health tasks while using 9x less test-time compute.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2506.10947","ref_index":3,"ref_count":1,"confidence":0.9,"is_internal_anchor":false,"paper_title":"Spurious Rewards: Rethinking Training Signals in RLVR","primary_cat":"cs.AI","submitted_at":"2025-06-12T17:49:55+00:00","verdict":"ACCEPT","verdict_confidence":"LOW","novelty_score":8.0,"formal_verification":"none","one_line_summary":"Spurious rewards in RLVR can produce large gains in mathematical reasoning for certain language models via GRPO's clipping bias amplifying pretraining behaviors like code reasoning.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null}],"limit":50,"offset":0}