QVal is a new evaluation framework that directly measures dense supervision quality via Q-alignment to a reference policy, showing simple prompting baselines outperform 21 other methods across environments and models.
hub
arXiv preprint arXiv:2602.02482 , year=
19 Pith papers cite this work. Polarity classification is still indexing.
19
Pith papers citing it
hub tools
citation-role summary
background 4
citation-polarity summary
years
2026 19roles
background 4polarities
background 4representative citing papers
OPD+ removes the bias from stop-gradient in on-policy distillation by deriving correct gradients for f-divergences, outperforming standard KL-based methods on math reasoning and tool-use tasks.
CORE distills contrasts between successful and unsuccessful reasoning traces into compact natural-language insights that enable faster model self-improvement on reasoning tasks with fewer rollouts than parametric or other non-parametric baselines.
VPD frames language feedback learning as variational EM so the teacher policy refines itself via trust-region updates on outcomes while the student learns dense token distributions on its own rollouts, outperforming fixed-teacher baselines on reasoning and code tasks.
RLRT augments GRPO by reinforcing tokens on correct student rollouts that the teacher would not have predicted, outperforming standard self-distillation and exploration baselines on Qwen3 models.
GUI-SD introduces on-policy self-distillation with visually enriched privileged context and entropy-guided weighting, outperforming GRPO and naive OPSD on six GUI grounding benchmarks while improving training efficiency.
AXPO addresses the Thinking-Acting Gap in agentic RL training by targeted resampling of tool calls in all-wrong subgroups, delivering +1.8pp gains over GRPO on nine multimodal benchmarks with an 8B model beating a 32B baseline on Pass@4.
ECHO is a hybrid RL objective that trains agents to predict environment observation tokens from their actions, doubling GRPO pass@1 on TerminalBench-2.0 while improving dynamics prediction on held-out trajectories.
DITTO uses RL with verbal feedback to train LLMs for human behavior simulation, reporting 36% average gains over base models and outperforming GPT-5.4 on 6 of 10 SOUL benchmark tasks.
NudgeRL conditions RLVR rollouts on strategy-level contexts to drive diverse trajectories and applies an inter/intra-context reward decomposition plus distillation objective, outperforming GRPO and oracle baselines on math benchmarks.
EchoDistill applies noisy-to-clean self-distillation with GRPO to boost Audio LLM robustness, reporting 4.18% average GSR gains under strong noise.
Prefix Sampling replays self-generated trajectory prefixes to control rollout pass rates near 50% in binary-reward RL, delivering wall-clock speedups and modest performance gains on SWE-bench Verified and AIME tasks.
FlexSQL reaches 65.4% on Spider2-Snow by allowing agents to flexibly explore schemas, generate diverse plans, choose SQL or Python execution, and apply two-tiered repair.
CoARS enables co-evolving recommender and user agents by using interaction-derived rewards and self-distilled credit assignment to internalize multi-turn feedback into model parameters, outperforming prior agentic baselines.
SelfEvo enables pretrained 4D perception models to self-improve on unlabeled videos via self-distillation, delivering up to 36.5% relative gains in video depth estimation and 20.1% in camera estimation across eight benchmarks.
A linear relationship between initial student-self-teacher performance gap and OPSD improvement provides a predictive law across contexts and model families.
Bi-NAC frames RL with textual feedback as a Stackelberg bilevel program and reports that 2B and 6B models trained this way outperform larger GRPO baselines on MATH-500 and GPQA.
SPEAR enables online federated LLM fine-tuning by using feedback-guided self-play to create contrastive pairs trained with maximum likelihood on correct completions and confidence-weighted unlikelihood on incorrect ones, outperforming baselines without ground-truth contexts.
PGPO modulates per-step trust in self-distilled updates via a mutual-information estimate derived from a viscous-fluid analogy, preserves SGD weak-approximation order, and reports gains of up to 4.5 points on Science-QA while avoiding late-training collapse.
citing papers explorer
-
QVal: Cheaply Evaluating Dense Supervision Signals for Long-Horizon LLM Agents
QVal is a new evaluation framework that directly measures dense supervision quality via Q-alignment to a reference policy, showing simple prompting baselines outperform 21 other methods across environments and models.
-
OPD+: Rethinking the Advantage Design for On-Policy Distillation
OPD+ removes the bias from stop-gradient in on-policy distillation by deriving correct gradients for f-divergences, outperforming standard KL-based methods on math reasoning and tool-use tasks.
-
CORE: Contrastive Reflection Enables Rapid Improvements in Reasoning
CORE distills contrasts between successful and unsuccessful reasoning traces into compact natural-language insights that enable faster model self-improvement on reasoning tasks with fewer rollouts than parametric or other non-parametric baselines.
-
Learning from Language Feedback via Variational Policy Distillation
VPD frames language feedback learning as variational EM so the teacher policy refines itself via trust-region updates on outcomes while the student learns dense token distributions on its own rollouts, outperforming fixed-teacher baselines on reasoning and code tasks.
-
Rebellious Student: Reversing Teacher Signals for Reasoning Exploration with Self-Distilled RLVR
RLRT augments GRPO by reinforcing tokens on correct student rollouts that the teacher would not have predicted, outperforming standard self-distillation and exploration baselines on Qwen3 models.
-
Learn where to Click from Yourself: On-Policy Self-Distillation for GUI Grounding
GUI-SD introduces on-policy self-distillation with visually enriched privileged context and entropy-guided weighting, outperforming GRPO and naive OPSD on six GUI grounding benchmarks while improving training efficiency.
-
Agent Explorative Policy Optimization for Multimodal Agentic Reasoning
AXPO addresses the Thinking-Acting Gap in agentic RL training by targeted resampling of tool calls in all-wrong subgroups, delivering +1.8pp gains over GRPO on nine multimodal benchmarks with an 8B model beating a 32B baseline on Pass@4.
-
ECHO: Terminal Agents Learn World Models for Free
ECHO is a hybrid RL objective that trains agents to predict environment observation tokens from their actions, doubling GRPO pass@1 on TerminalBench-2.0 while improving dynamics prediction on held-out trajectories.
-
Reinforcing Human Behavior Simulation via Verbal Feedback
DITTO uses RL with verbal feedback to train LLMs for human behavior simulation, reporting 36% average gains over base models and outperforming GPT-5.4 on 6 of 10 SOUL benchmark tasks.
-
Nudging Beyond the Comfort Zone: Efficient Strategy-Guided Exploration for RLVR
NudgeRL conditions RLVR rollouts on strategy-level contexts to drive diverse trajectories and applies an inter/intra-context reward decomposition plus distillation objective, outperforming GRPO and oracle baselines on math benchmarks.
-
EchoDistill:Alignment Noisy-to-Clean Self-Distillation for Robust Audio LLMs
EchoDistill applies noisy-to-clean self-distillation with GRPO to boost Audio LLM robustness, reporting 4.18% average GSR gains under strong noise.
-
Rollout Pass-Rate Control: Steering Binary-Reward RL Toward Its Most Informative Regime
Prefix Sampling replays self-generated trajectory prefixes to control rollout pass rates near 50% in binary-reward RL, delivering wall-clock speedups and modest performance gains on SWE-bench Verified and AIME tasks.
-
FlexSQL: Flexible Exploration and Execution Make Better Text-to-SQL Agents
FlexSQL reaches 65.4% on Spider2-Snow by allowing agents to flexibly explore schemas, generate diverse plans, choose SQL or Python execution, and apply two-tiered repair.
-
Self-Distilled Reinforcement Learning for Co-Evolving Agentic Recommender Systems
CoARS enables co-evolving recommender and user agents by using interaction-derived rewards and self-distilled credit assignment to internalize multi-turn feedback into model parameters, outperforming prior agentic baselines.
-
Self-Improving 4D Perception via Self-Distillation
SelfEvo enables pretrained 4D perception models to self-improve on unlabeled videos via self-distillation, delivering up to 36.5% relative gains in video depth estimation and 20.1% in camera estimation across eight benchmarks.
-
A Predictive Law for On-Policy Self-Distillation From World Feedback
A linear relationship between initial student-self-teacher performance gap and OPSD improvement provides a predictive law across contexts and model families.
-
RL with Learnable Textual Feedback: A Bilevel Approach
Bi-NAC frames RL with textual feedback as a Stackelberg bilevel program and reports that 2B and 6B models trained this way outperform larger GRPO baselines on MATH-500 and GPQA.
-
Self-Play Enhancement via Advantage-Weighted Refinement in Online Federated LLM Fine-Tuning with Real-Time Feedback
SPEAR enables online federated LLM fine-tuning by using feedback-guided self-play to create contrastive pairs trained with maximum likelihood on correct completions and confidence-weighted unlikelihood on incorrect ones, outperforming baselines without ground-truth contexts.
-
Physics-Guided Policy Optimization with Self-Distillation
PGPO modulates per-step trust in self-distilled updates via a mutual-information estimate derived from a viscous-fluid analogy, preserves SGD weak-approximation order, and reports gains of up to 4.5 points on Science-QA while avoiding late-training collapse.