Auxiliary Rewards Improve Small-Language-Model Self-Improvement
UID-Conj. auxiliary rewards boost online adaptation and offline SFT by 2.08 and 1.93 points over baselines on AIME math tasks.
Stories per day · 30 days
Relatedactivation-steering1auxiliary-rewards1hmmt1reasoning-models1self-distillation1self-improvement1small-language-models1transfer-learning1
13 stories, newest first · one card per story
Stories tagged #aime
Steered-teacher self-distillation cuts reference hallucinations in thinking models
Activation steering is applied only to the teacher; the student is not steered, and AIME and HMMT accuracy rises over vanilla OPSD.
From Pareto to Preference: Personalized Test-Time Scaling via Amortized Agentic Policy Discovery
PersonTTS reuses prior search to find test-time controllers matching each user's accuracy, latency, and cost goals.
Semifactual Credit-Augmented Policy Optimization
SCAPO uses semifactual token stability to beat GRPO on AIME for two Qwen3 base models.
Semifactual Credit-Augmented Policy Optimization
SCAPO adds semifactual token stability to GRPO, lifting Qwen3 AIME accuracy by up to 5.63 points.
Better Supervision Is Nearby: Neighborhood On-Policy Self-Distillation
Neighborhood OPSD lifts Qwen3 math scores by routing student prefixes to nearby expert teachers.
EasyPPO: Stabilizing the Critic Is Key
EasyPPO stabilizes the PPO critic for LLM training and outperforms vanilla PPO on coding, math, and search.
Overcoming Scaling Limits in On-Policy Self-Distillation for LLM Reasoning
OASIS keeps on-policy self-distillation effective at scale by supervising verified reasoning trajectories.
Improving Test-Time Scaling with Adaptive Looped Transformers
TaH2 adaptively loops tokens that benefit, steepening the AIME accuracy-compute slope 53 percent versus a non-looped baseline.
Improving Test-Time Scaling with Adaptive Looped Transformers
TaH2 adaptively loops layers on hard tokens, steepening AIME test-time scaling by 53% over a non-looped baseline.
Not Every Token Is Worth Distilling: Selective Supervision for Direct-OPD
Selective Direct-OPD keeps the top 10% of high-divergence states and beats dense distillation on math benchmarks.
Self-Organizing Agent Teams Learn to Reason Together
Self-organizing agent teams learn collaboration strategies and beat their strongest member on math and physics benchmarks.
When2Think: Learning Difficulty-Aware Length Control for Efficient Hybrid Reasoning Models
When2Think post-training framework allocates reasoning depth by problem difficulty, raising AIME24 Pass@3 10.0% while cutting token usage 27.9%.