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.
hub
Can large reasoning models self-train?
20 Pith papers cite this work. Polarity classification is still indexing.
20
Pith papers citing it
hub tools
citation-role summary
background 3
method 1
citation-polarity summary
representative citing papers
PACE is a training-free anytime-valid commit gate using testing-by-betting e-processes that controls per-candidate false-commit probability for self-evolving agents and reduces spurious edits compared to greedy acceptance.
TTRL-CoCoV is a confidence-conditioned test-time RL framework that selectively applies verification to address pseudo-label errors and diversity collapse, yielding +9.8% Pass@1 and +18.7% Pass@16 gains over prior TTRL on reasoning benchmarks.
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.
SFT supplies entangled compositional traces of atomic skills and routing modules; RL identifies those modules and enables recombination on novel compositions outside the SFT support.
REINFORCE self-training on competitive programming tasks exhibits robust rise-then-collapse in pass@1; CARE, ES, and GRPO mitigate it in model-size-dependent ways across Qwen-2.5-3B/7B and a Gemma pilot.
Unsupervised rewards combining model uncertainty and semantic consistency allow protein language models to self-steer via SRO and BRO algorithms, outperforming DPO and KTO on out-of-distribution prompts while approaching oracle performance.
Experiments on coding and deterministic tasks demonstrate that data gating is sufficient for self-play stability while reward variants are not, revealing the Grounded Proposer Paradox and a two-stage phase transition under continuous gate strictness.
G-Zero uses the Hint-δ intrinsic reward to drive co-evolution between a Proposer and Generator via GRPO and DPO, providing a theoretical suboptimality guarantee for self-improvement from internal dynamics alone.
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.
SimpleTES scales test-time evaluation in LLMs to discover state-of-the-art solutions on 21 scientific problems across six domains, outperforming frontier models and optimization pipelines with examples like 2x faster LASSO and new Erdos constructions.
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.
Co-evolving coder and tester models via consensus over a self-generated pass matrix improves LLM code generation up to 14.5% label-free and 21.6% with a lightly calibrated Bayesian selector.
ETS performs training-free RL alignment for language models by energy-guided test-time scaling with Monte Carlo energy estimation and importance sampling acceleration.
CoNL lets LLMs self-improve on non-verifiable tasks by rewarding critiques that produce better solutions in multi-agent conversations, jointly optimizing generation and judging without external feedback.
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.
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.
Closed-loop self-evolution on LLMs improves reasoning on Knights and Knaves tasks but plateaus short of oracle-supervised levels, with multi-turn revision nearly matching it for large models.
This survey frames foundation agents using brain-inspired modular architectures and reviews challenges in evolution, collaboration, and safety.
citing papers explorer
-
When Do Intrinsic Rewards Work for Code Reasoning? A Comprehensive Study
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.
-
PACE: Anytime-Valid Acceptance Tests for Self-Evolving Agents
PACE is a training-free anytime-valid commit gate using testing-by-betting e-processes that controls per-candidate false-commit probability for self-evolving agents and reduces spurious edits compared to greedy acceptance.
-
Exploiting Verification-Generation Gap: Test-Time Reinforcement Learning with Confidence-Conditioned Verification
TTRL-CoCoV is a confidence-conditioned test-time RL framework that selectively applies verification to address pseudo-label errors and diversity collapse, yielding +9.8% Pass@1 and +18.7% Pass@16 gains over prior TTRL on reasoning benchmarks.
-
What If Consensus Lies? Selective-Complementary Reinforcement Learning at Test Time
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.
-
From Reasoning Traces to Reusable Modules: Understanding Compositional Generalization in Language Model Reasoning
SFT supplies entangled compositional traces of atomic skills and routing modules; RL identifies those modules and enables recombination on novel compositions outside the SFT support.
-
Self-Improvement Can Self-Regress: The Rise-and-Collapse Failure Mode of LLM Self-Training
REINFORCE self-training on competitive programming tasks exhibits robust rise-then-collapse in pass@1; CARE, ES, and GRPO mitigate it in model-size-dependent ways across Qwen-2.5-3B/7B and a Gemma pilot.
-
Be Your Own Teacher: Steering Protein Language Models via Unsupervised Reward Optimization
Unsupervised rewards combining model uncertainty and semantic consistency allow protein language models to self-steer via SRO and BRO algorithms, outperforming DPO and KTO on out-of-distribution prompts while approaching oracle performance.
-
Survive or Collapse: The Asymmetric Roles of Data Gating and Reward Grounding in Self-Play RL
Experiments on coding and deterministic tasks demonstrate that data gating is sufficient for self-play stability while reward variants are not, revealing the Grounded Proposer Paradox and a two-stage phase transition under continuous gate strictness.
-
G-Zero: Self-Play for Open-Ended Generation from Zero Data
G-Zero uses the Hint-δ intrinsic reward to drive co-evolution between a Proposer and Generator via GRPO and DPO, providing a theoretical suboptimality guarantee for self-improvement from internal dynamics alone.
-
Experience Sharing in Mutual Reinforcement Learning for Heterogeneous Language Models
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.
-
Evaluation-driven Scaling for Scientific Discovery
SimpleTES scales test-time evaluation in LLMs to discover state-of-the-art solutions on 21 scientific problems across six domains, outperforming frontier models and optimization pipelines with examples like 2x faster LASSO and new Erdos constructions.
-
Too Correct to Learn: Reinforcement Learning on Saturated Reasoning Data
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.
-
ZeroCoder: Can LLMs Improve Code Generation Without Ground-Truth Supervision?
Co-evolving coder and tester models via consensus over a self-generated pass matrix improves LLM code generation up to 14.5% label-free and 21.6% with a lightly calibrated Bayesian selector.
-
ETS: Energy-Guided Test-Time Scaling for Training-Free RL Alignment
ETS performs training-free RL alignment for language models by energy-guided test-time scaling with Monte Carlo energy estimation and importance sampling acceleration.
-
Conversation for Non-verifiable Learning: Self-Evolving LLMs through Meta-Evaluation
CoNL lets LLMs self-improve on non-verifiable tasks by rewarding critiques that produce better solutions in multi-agent conversations, jointly optimizing generation and judging without external feedback.
-
GeoMin: Data-Efficient Semi-Supervised RLVR via Geometric Distribution Modeling
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.
-
Trust Region On-Policy Distillation
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.
-
On the Generalization Gap in Self-Evolving Language Model Reasoning
Closed-loop self-evolution on LLMs improves reasoning on Knights and Knaves tasks but plateaus short of oracle-supervised levels, with multi-turn revision nearly matching it for large models.
-
Advances and Challenges in Foundation Agents: From Brain-Inspired Intelligence to Evolutionary, Collaborative, and Safe Systems
This survey frames foundation agents using brain-inspired modular architectures and reviews challenges in evolution, collaboration, and safety.
- MetaLint: Easy-to-Hard Generalization for Code Linting