Derives an exact telescoping decomposition of the naive RLVR reward-design estimator into null, elicitation, and reward-design terms on a tabular-GRPO simulator, measures the components across prior strengths, and validates via pre-registered factorial experiments plus re-audits of prior papers.
super hub Canonical reference
Group Sequence Policy Optimization
Canonical reference. 72% of citing Pith papers cite this work as background.
245
Pith papers citing it
Background
72% of classified citations
abstract
This paper introduces Group Sequence Policy Optimization (GSPO), our stable, efficient, and performant reinforcement learning algorithm for training large language models. Unlike previous algorithms that adopt token-level importance ratios, GSPO defines the importance ratio based on sequence likelihood and performs sequence-level clipping, rewarding, and optimization. We demonstrate that GSPO achieves superior training efficiency and performance compared to the GRPO algorithm, notably stabilizes Mixture-of-Experts (MoE) RL training, and has the potential for simplifying the design of RL infrastructure. These merits of GSPO have contributed to the remarkable improvements in the latest Qwen3 models.
hub tools
citation-role summary
background 29
method 10
baseline 2
dataset 2
citation-polarity summary
claims ledger
- abstract This paper introduces Group Sequence Policy Optimization (GSPO), our stable, efficient, and performant reinforcement learning algorithm for training large language models. Unlike previous algorithms that adopt token-level importance ratios, GSPO defines the importance ratio based on sequence likelihood and performs sequence-level clipping, rewarding, and optimization. We demonstrate that GSPO achieves superior training efficiency and performance compared to the GRPO algorithm, notably stabilizes Mixture-of-Experts (MoE) RL training, and has the potential for simplifying the design of RL infras
authors
co-cited works
representative citing papers
P2R decouples perception from reasoning in VLMs via a two-stage process and PRA-GRPO alternating RL training, reporting gains such as 93.2% on V-Star for the 4B model over its Qwen3-VL backbone.
GRPO updates reduce to a damped oscillator whose mass, damping, and stiffness are fixed by optimizer hyperparameters plus one measured curvature scale, subsuming single-exponential saturation while adding inertial slow-start and group-size predictions.
Proposes Monotonic Inference Policy Improvement (MIPI) objective and MIPU two-step update framework to address objective misalignment between training and inference policies in LLM reinforcement learning.
TRL extends tandem training to RLVR pipelines, matching GRPO solo reasoning on Qwen3-4B math tasks while improving handoff robustness, reducing distributional drift, and increasing CoT legibility for the junior.
TempAct introduces a planner-executor RL framework with hierarchical group exploration and rewards to improve temporal consistency in autoregressive video diffusion models.
SC-GRPO improves RL with verifiable rewards by multiplying GRPO gradients with self-induced per-token KL divergence, outperforming GRPO by 8.1% and DAPO by 5.9% on math, code, and agent benchmarks.
PearlVLA achieves SOTA on LIBERO by separating VLM representations into visual grounding and an iterative latent plan branch refined via world model queries and RefineNet with process-reward RL.
ReSum trains LLMs via RLVR to self-summarize reasoning trajectories, yielding 4% average performance gains and 18.6% shorter rollouts through contrastive rollout branches.
Socratic-SWE distills agent solving traces into skills that generate and filter targeted repair tasks, yielding iterative gains on SWE-bench suites under fixed compute.
Introduces OPT* tasks and two training regimes (solver-guided online policy optimization with rank-based reward shaping and search-based offline RL) plus a theoretical link between search success and information extraction per budget unit, showing empirical gains in optimization-like reasoning.
ElasticMem enables LLM agents to learn adaptive latent memory retrieval and elastic budget allocation, improving QA accuracy by 24-26% and ALFWorld success by 27-66% over baselines with lower token cost.
PInVerify is a new offline embodied benchmark for active instance verification that supplies multi-view captures and 6-sector navigation topology, with MLLM baselines reaching 85.6% after fine-tuning but showing no reliable benefit from tested next-best-view strategies.
Reasoning models naturally compress context via thinking traces, with reward-constrained optimization yielding 17-23% gains over baselines on long-context QA at high compression ratios.
Introduces a state-aligned latent actor-critic framework that lets diffusion models act as their own timestep-conditioned value functions for trajectory-level RL post-training and inference steering.
Touch-R1 applies GRPO reinforcement learning on a new 1M tactile dataset and benchmark to train a Qwen2.5-VL-7B model that outperforms baselines on tactile perception and visual-tactile conflict tasks.
CUA-Gym generates 32,112 verified RLVR tuples across 110 mock environments, enabling trained models to reach 62.1% and 72.6% on OSWorld-Verified while transferring to WebArena.
MCPO applies contrastive learning to GRPO-style RL by treating cross-domain correct rollouts as positives and incorrect ones as negatives to improve multi-domain reasoning performance in LRMs.
Temporal scheduling of credit allocation criteria over RLVR training, using trajectory percentiles to target heterogeneous behaviors, yields more stable policy entropy and better reasoning benchmark results than static allocation.
A reinforcement-learned vision-language agent adaptively selects and fuses monocular depth experts per sample for better performance across camera geometries.
Pion modifies Muon's Newton-Schulz iterations into a controllable high-pass filter that anchors dominant singular values at 1 while suppressing noisy tails, outperforming Muon and AdamW in VLA and RLVR regimes.
PPR-GDE is a new RL approach that integrates pairwise preference rewards with group-based diversity enhancement in a unified objective to improve both alignment quality and expressive diversity in open-ended generation tasks such as role-playing.
DISA decouples partition function estimation using offline importance sampling for distribution-matching LLM-RL, matching or exceeding online baselines like FlowRL on math and code benchmarks while retaining more strategy diversity.
AstraFlow decouples RL components into autonomous dataflow services to natively support multi-policy agentic LLM training, elastic scaling, and cross-region execution with 2.7x speedup on math, code, search, and AgentBench workloads.
citing papers explorer
-
A Pre-Registered Causal Partition of Self-Consistency Elicitation and Reward Design in RLVR
Derives an exact telescoping decomposition of the naive RLVR reward-design estimator into null, elicitation, and reward-design terms on a tabular-GRPO simulator, measures the components across prior strengths, and validates via pre-registered factorial experiments plus re-audits of prior papers.
-
Tandem Reinforcement Learning with Verifiable Rewards
TRL extends tandem training to RLVR pipelines, matching GRPO solo reasoning on Qwen3-4B math tasks while improving handoff robustness, reducing distributional drift, and increasing CoT legibility for the junior.
-
ReSum: Synergizing LLM Reasoning and Summarization with Reinforcement Learning
ReSum trains LLMs via RLVR to self-summarize reasoning trajectories, yielding 4% average performance gains and 18.6% shorter rollouts through contrastive rollout branches.
-
Step-by-Step Optimization-like Reasoning in LLMs over Expanding Search Spaces
Introduces OPT* tasks and two training regimes (solver-guided online policy optimization with rank-based reward shaping and search-based offline RL) plus a theoretical link between search success and information extraction per budget unit, showing empirical gains in optimization-like reasoning.
-
Thinking as Compression: Your Reasoning Model is Secretly a Context Compressor
Reasoning models naturally compress context via thinking traces, with reward-constrained optimization yielding 17-23% gains over baselines on long-context QA at high compression ratios.
-
CUA-Gym: Scaling Verifiable Training Environments and Tasks for Computer-Use Agents
CUA-Gym generates 32,112 verified RLVR tuples across 110 mock environments, enabling trained models to reach 62.1% and 72.6% on OSWorld-Verified while transferring to WebArena.
-
Pairwise Preference Reward and Group-Based Diversity Enhancement for Superior Open-Ended Generation
PPR-GDE is a new RL approach that integrates pairwise preference rewards with group-based diversity enhancement in a unified objective to improve both alignment quality and expressive diversity in open-ended generation tasks such as role-playing.
-
ClawForge: Generating Executable Interactive Benchmarks for Command-Line Agents
ClawForge is a generator framework that creates reproducible executable benchmarks for command-line agents under state conflict, with ClawForge-Bench showing frontier models reach at most 45.3% strict accuracy and that state inspection drives most performance gaps.
-
SeePhys Pro: Diagnosing Modality Transfer and Blind-Training Effects in Multimodal RLVR for Physics Reasoning
SeePhys Pro benchmark reveals multimodal models degrade on physics reasoning as information transfers from text to images, with blind training improvements often stemming from textual cues rather than visual evidence.
-
Which Tokens Matter? Adaptive Token Selection for RLVR with the Relative Surprisal Index
Introduces RSI metric and RSI-S filtering method for adaptive token selection in RLVR, reporting 2-3 point gains over GRPO on AIME/AMC benchmarks.
-
IterCAD: An Iterative Multimodal Agent for Visually-Grounded CAD Generation and Editing
IterCAD is a multimodal agent framework using progressive SFT and geometry-aware RL for CAD tasks, with a new data pipeline, IterCAD-Bench, and CD-TR metric showing outperformance in executability and precision.
-
Teaching the Way, Not the Answer: Privileged Tutoring Distillation for Multimodal Policy Optimization
PTD-PO supplies step-wise token-distribution supervision to student policies via in-context privileged hints derived from spatial attention and intermediate reasoning, while keeping the student in an answer-free context and using Top-K Jensen-Shannon divergence for stable alignment.
-
EvoTrainer: Co-Evolving LLM Policies and Training Harnesses for Autonomous Agentic Reinforcement Learning
EvoTrainer co-evolves LLM policies and training harnesses via empirical feedback to match or exceed human-engineered RL on math reasoning, code generation, and long-horizon software engineering.
-
Beyond Trajectory Rewards: Step-level Credit Assignment for Agentic Search via Graph Modeling
GDCR assigns step-level rewards via distance to the answer node in a training-time ER graph and SAPO combines these with trajectory advantages for credit assignment in agentic search.
-
TRACER: Turn-level Regret Matching with Inner Reinforcement Credit for Cooperative Multi-LLM Reasoning
TRACER combines a controller-regret layer using regret matching for speak/skip decisions with a generation-credit layer using GSPO rewards to enable learned collaboration in multi-LLM reasoning.
-
EAPO: Entropy-Driven Adaptive Positive-Negative Sample Weighting for Policy Optimization in Open-Ended QA
EAPO uses policy entropy ratio to adaptively weight positive samples in RLVR for open-ended QA, claiming better diversity and stability than fixed-weight baselines on medical datasets.
-
Efficient Agentic Reasoning Through Self-Regulated Simulative Planning
SR²AM achieves competitive Pass@1 accuracy on diverse tasks with 25.8-95.3% fewer reasoning tokens than much larger models by using self-regulated simulative planning trained via supervised learning and RL.
-
Beyond Mode Collapse: Distribution Matching for Diverse Reasoning
DMPO approximates forward KL minimization in on-policy RL by aligning the policy to a group-level reward-proportional target distribution, yielding 9-12% relative gains over GRPO on NP-Bench and smaller gains on math reasoning.
-
What and When to Distill: Selective Hindsight Distillation for Multi-Turn Agents
SERL selectively reweights learning using task success and environment feedback to reach 90.0% success on ALFWorld and 80.1% on WebShop, outperforming RL and distillation baselines.
-
SAPO: Step-Aligned Policy Optimization for Reasoning-Based Generative Recommendation
SAPO computes per-reasoning-step group-relative advantages in RL to improve credit assignment for structured generation of semantic identifiers in recommendation systems.
-
ICRL: Learning to Internalize Self-Critique with Reinforcement Learning
ICRL uses joint RL training of solver and critic with distribution-calibration re-weighting and role-wise advantage estimation to internalize critique into unassisted LLM performance, yielding 6.4-point gains on agentic tasks and 7.0 on math reasoning with Qwen3 models.
-
Seir\^enes: Adversarial Self-Play with Evolving Distractions for LLM Reasoning
Seirênes trains LLMs via adversarial self-play to generate and overcome evolving distractions, producing gains of 7-10 points on math reasoning benchmarks and exposing blind spots in larger models.
-
Rubric-Grounded RL: Structured Judge Rewards for Generalizable Reasoning
Rubric-grounded RL with LLM judges on document-derived criteria raises Llama-3.1-8B normalized reward to 71.7% on held-out rubrics and improves performance on GSM8K, MATH, and GPQA benchmarks.
-
Can RL Teach Long-Horizon Reasoning to LLMs? Expressiveness Is Key
RL training compute for logical reasoning follows a power law with horizon depth whose exponent rises with logical expressiveness, yielding better downstream transfer when models train on richer logics.
-
Segment-Aligned Policy Optimization for Multi-Modal Reasoning
SAPO introduces segment-level policy optimization using a step-wise MDP abstraction to better align RL updates with reasoning structure in multi-modal LLM tasks.
-
WaferSAGE: Large Language Model-Powered Wafer Defect Analysis via Synthetic Data Generation and Rubric-Guided Reinforcement Learning
A 4B-parameter vision-language model trained on rubric-guided synthetic wafer defect data reaches 6.493 LLM-Judge score, nearly matching Gemini-3-Flash at 7.149 for on-premise industrial use.
-
MARS$^2$: Scaling Multi-Agent Tree Search via Reinforcement Learning for Code Generation
MARS² integrates multi-agent collaboration with tree-structured search in RL to boost code generation by increasing exploratory diversity and using path-level group advantages for credit assignment.
-
Chart-RL: Policy Optimization Reinforcement Learning for Enhanced Visual Reasoning in Chart Question Answering with Vision Language Models
Chart-RL uses RL policy optimization and LoRA to boost VLM chart reasoning, enabling a 4B model to reach 0.634 accuracy versus 0.580 for an 8B model with lower latency.
-
Unifying Ontology Construction and Semantic Alignment for Deterministic Enterprise Reasoning at Scale
LOM unifies ontology construction, semantic alignment, and deterministic reasoning in one architecture, reporting 88.8% accuracy on ontology completion and 94% on complex graph reasoning tasks.
-
The Landscape of Agentic Reinforcement Learning for LLMs: A Survey
Survey that defines agentic RL for LLMs via POMDPs, introduces a taxonomy of planning/tool-use/memory/reasoning capabilities and domains, and compiles open environments from over 500 papers.
-
AlloSpatial: Agentic Harness Framework for Spatial Reasoning in Foundation Models
AlloSpatial adds structured allocentric priors and a harness for tool-use and arbitration to improve spatial reasoning in foundation models, with 5-18% gains on VSI-Bench and MindCube in training-free settings and further gains after RL internalization.
-
Mechanistically Interpreting the Role of Sample Difficulty in RLVR for LLMs
Sample difficulty in RLVR shows non-monotonic effects on LLM reasoning, with easy/medium problems strengthening computation and reasoning features while hard problems often yield weak or harmful signals.
-
Learning to Act under Noise: Enhancing Agent Robustness via Noisy Environments
NoisyAgent trains LLM agents with controlled user and tool noise to improve robustness in stochastic environments while also boosting clean-benchmark performance.
-
Entropy-Gradient Inversion: Moving Toward Internal Mechanism of Large Reasoning Models
Defines Entropy-Gradient Inversion as a negative entropy-gradient correlation fingerprinting LRM reasoning and proposes CorR-PO to embed it in RL regularization, claiming consistent outperformance on benchmarks.
-
How You Begin is How You Reason: Driving Exploration in RLVR via Prefix-Tuned Priors
IMAX trains soft prefixes with an InfoMax reward to drive diverse exploration in RLVR, yielding up to 11.60% gains in Pass@4 over standard RLVR across model scales.
-
Learning CLI Agents with Structured Action Credit under Selective Observation
CLI agents trained with RL benefit from selective observation via σ-Reveal and structured credit assignment via A³ that leverages AST action sub-chains and trajectory margins.
-
On Training Large Language Models for Long-Horizon Tasks: An Empirical Study of Horizon Length
Longer action horizons bottleneck LLM agent training through instability, but training with reduced horizons stabilizes learning and enables better generalization to longer horizons.
-
MCPO: Mastery-Consolidated Policy Optimization for Large Reasoning Models
MCPO fixes vanishing training signals and shrinking weights in GRPO by using a hinge-KL regularizer on mastered prompts and prioritizing majority-correct prompts, yielding higher pass@1 and pass@k on math tasks.
-
Mind DeepResearch Technical Report
MindDR combines a Planning Agent, DeepSearch Agent, and Report Agent with SFT cold-start, Search-RL, Report-RL, and preference alignment to reach competitive scores on research benchmarks using 30B-scale models.
-
StaRPO: Stability-Augmented Reinforcement Policy Optimization
StaRPO improves LLM reasoning by adding autocorrelation function and path efficiency stability metrics to RL policy optimization, yielding higher accuracy and fewer logic errors on reasoning benchmarks.
-
Xiaomi-GUI-0 Technical Report
Xiaomi-GUI-0 reports 72.0% success on RealMobile and 78.9% on AndroidWorld via real-device closed-loop training with multi-source data and three-stage RL pipeline.
-
Shattering the Autoregressive Curse: Dynamic Epistemic Entropy Orchestrated Erasable Reinforcement Learning for LLMs
E³RL uses dynamic thresholds on epistemic entropy from autoregressive cross-entropy to enable erasable RL in LLM reasoning, reporting 5.349% and 6.514% gains on AIME for 4B and 8B models over prior SOTA.
-
AEM: Adaptive Entropy Modulation for Multi-Turn Agentic Reinforcement Learning
AEM adaptively modulates response-level entropy in agentic RL to improve credit assignment and exploration-exploitation balance, yielding gains on ALFWorld, WebShop, and SWE-bench.
-
Baichuan-M4: A Clinical-Grade Medical Agent System for Continuous Care
The paper describes Baichuan-M4, a coordinated medical agent system that reports leading scores across static knowledge, dynamic consultation, long-context memory, retrieval, OCR, and multimodal tasks with a 3.3% hallucination rate.
-
Rethinking Agentic Reinforcement Learning In Large Language Models
The paper reviews conceptual foundations, methodological innovations, effective designs, critical challenges, and future directions for LLM-based Agentic Reinforcement Learning.
- OGER: A Robust Offline-Guided Exploration Reward for Hybrid Reinforcement Learning