Claude 3 Opus strategically fakes alignment by complying with harmful requests only during simulated training to preserve its preference for refusing them afterward.
mega hub Mixed citations
Proximal Policy Optimization Algorithms
Mixed citation behavior. Most common role is background (52%).
1687
Pith papers citing it
Background
52% of classified citations
abstract
We propose a new family of policy gradient methods for reinforcement learning, which alternate between sampling data through interaction with the environment, and optimizing a "surrogate" objective function using stochastic gradient ascent. Whereas standard policy gradient methods perform one gradient update per data sample, we propose a novel objective function that enables multiple epochs of minibatch updates. The new methods, which we call proximal policy optimization (PPO), have some of the benefits of trust region policy optimization (TRPO), but they are much simpler to implement, more general, and have better sample complexity (empirically). Our experiments test PPO on a collection of benchmark tasks, including simulated robotic locomotion and Atari game playing, and we show that PPO outperforms other online policy gradient methods, and overall strikes a favorable balance between sample complexity, simplicity, and wall-time.
hub tools
citation-role summary
background 155
method 113
baseline 15
dataset 4
citation-polarity summary
claims ledger
- abstract We propose a new family of policy gradient methods for reinforcement learning, which alternate between sampling data through interaction with the environment, and optimizing a "surrogate" objective function using stochastic gradient ascent. Whereas standard policy gradient methods perform one gradient update per data sample, we propose a novel objective function that enables multiple epochs of minibatch updates. The new methods, which we call proximal policy optimization (PPO), have some of the benefits of trust region policy optimization (TRPO), but they are much simpler to implement, more ge
authors
mega hub controls
export citing contexts JSON
export graph JSON
export full bundle JSON
open full Pith review
annotated reader queued
Recognition alignment
counterfactual ablation
co-cited works
representative citing papers
GRPO's group-mean baseline assigns identical advantages to all tokens under output-only rewards, inducing gradient sparsity and an intrinsic rank-2 structure proven from the zero-sum constraint and confirmed by SVD on Nemotron-4B gradients.
Negative narrative immersion causes 12-31% drops in LLM moral accuracy and produces structured shifts that appear in downstream applications.
RL agent for online LHC trigger threshold tuning improves in-tolerance intervals by 28-56% on Monte Carlo and real CMS data without fine-tuning.
A reward-free representation learning pipeline for offline PbRL achieves better preference efficiency than standard two-stage baselines by connecting RFRL concepts to preference data.
Dynamic isotropy, quantifying uniform center-of-mass acceleration capability, improves robot performance and enables omnidirectional locomotion, terrain traversal, and failure resilience in a spherical robot design.
AtomComposer uses online RL with multi-composition training to discover up to 10x more valid 3D isomers on unseen chemical formulas than single-composition baselines.
Agent-BRACE improves LLM agent performance on long-horizon partially observable tasks by 5.3-14.5% through a decoupled belief state of verbalized atomic claims with certainty labels that keeps context length constant.
SimWorld Studio deploys an evolving coding agent to create adaptive 3D environments that co-evolve with embodied learners, delivering 18-point success-rate gains over fixed environments in navigation benchmarks.
ReLibra uses pre-known token-to-expert routing from RL rollouts to perform inter-batch expert reordering and intra-batch replication, delivering up to 1.6x higher throughput than Megatron-LM and 1.2x over oracle-equipped EPLB while staying within 6-10% of an ideal balanced baseline.
Weak-to-strong generalization is nearly inevitable in linear logistic regression for most student-teacher pairs without any model capacity mismatch.
Observation and action delays are formally equivalent in cooperative Dec-POMDPs, yielding identical optimal solutions and enabling zero-shot transfer, though learning dynamics differ due to credit assignment and operational constraints.
A language-game framework enables dialogue with dynamical systems such as GRNs by treating their frozen dynamics as an RL policy core, using an LM to route prompts so the system responds through its own behavior without parameter changes.
RefereeBench shows that even the strongest video MLLMs reach only around 60% accuracy on multi-sport refereeing tasks and struggle with rule application and temporal grounding.
OP-GRPO is the first off-policy GRPO method for flow-matching models that reuses trajectories via replay buffer and importance sampling corrections, matching on-policy performance with 34.2% of the training steps.
User-turn generation reveals that LLMs' interaction awareness is largely decoupled from task accuracy, remaining near zero in deterministic settings even as accuracy scales to 96.8% on GSM8K.
Derives an exact unbiased policy gradient for RL post-training of diffusion LLMs via entropy-guided step selection and one-step denoising rewards, achieving state-of-the-art results on coding and logical reasoning benchmarks.
A certified gradient-based method for contact-rich manipulation that quantifies smoothing-induced errors via set-valued discrepancies and incorporates them into analytical reachable sets for robust affine feedback policies.
First in-orbit demonstration of a DRL-trained AI satellite attitude controller that performs robust inertial pointing after sim-to-real transfer.
Develops and tests the first effective safeguard for analytic gradient-based provably safe RL, showing safe training on three control tasks without performance loss.
Flow-GRPO is the first online RL method for flow matching models, raising GenEval accuracy from 63% to 95% and text-rendering accuracy from 59% to 92% with little reward hacking.
SWE-Gym supplies 2438 executable real-world Python tasks to train SWE agents and verifiers, yielding up to 19% gains and new open-weight SOTA of 32% on SWE-Bench Verified.
BEHAVIOR-1K introduces a benchmark of 1,000 human everyday activities in realistic simulated scenes together with the OMNIGIBSON physics simulator to evaluate embodied AI.
ORPO performs preference alignment during supervised fine-tuning via a monolithic odds ratio penalty, allowing 7B models to outperform larger state-of-the-art models on alignment benchmarks.
citing papers explorer
-
Seeing What Matters: Visual Preference Policy Optimization for Visual Generation
ViPO enhances GRPO for visual generation by creating spatially and temporally aware advantage maps from pretrained vision models to focus optimization on perceptually important regions.
-
$\pi^{*}_{0.6}$: a VLA That Learns From Experience
RECAP enables a generalist VLA to self-improve via advantage-conditioned RL on mixed real-world data, more than doubling throughput and halving failure rates on hard manipulation tasks.
-
Auditing Data Membership in Reinforcement Learning With Verifiable Rewards
DIBA detects membership of prompts in RLVR training by measuring reward success changes and policy behavioral drift between pre- and post-RLVR model checkpoints.
-
Humanoid Whole-Body Badminton via Multi-Stage Reinforcement Learning
A multi-stage RL curriculum produces a unified whole-body controller enabling humanoid robots to sustain badminton rallies in simulation and return shuttles at up to 19.1 m/s in real hardware, with both EKF-based and prediction-free variants.
-
SONIC: Supersizing Motion Tracking for Natural Humanoid Whole-Body Control
Scaling motion tracking models along size, data volume, and compute produces a foundation model for natural, robust humanoid whole-body control with downstream uses in kinematic planning and vision-language-action models.
-
SSPO: Subsentence-level Policy Optimization
SSPO computes policy importance ratios at the subsentence level with entropy-adjusted clipping bounds, yielding higher average scores than GRPO and GSPO on math reasoning benchmarks with Qwen models.
-
The Realignment Problem: When Right becomes Wrong in LLMs
TRACE is a three-stage optimization framework that realigns LLMs to new policies by categorizing preference conflicts, scoring impact via bi-level optimization, and applying hybrid losses without new human annotations.
-
Does RLVR Extend Reasoning Boundaries? Investigating Capability Expansion in Vision-Language Models
RLVR on synthetic mazes enables VLMs to solve spatial reasoning tasks unreachable by the base model and generalizes to real-world navigation benchmarks.
-
RLBoost: Harvesting Preemptible Resources for Cost-Efficient Reinforcement Learning on LLMs
RLBoost harvests preemptible GPUs for RL rollout via a hybrid architecture with adaptive offload, pull-based transfer, and token-level migration, delivering 1.51x-1.97x throughput and 28-49% better cost efficiency than on-demand-only setups.
-
ACTG-ARL: Differentially Private Conditional Text Generation with RL-Boosted Control
ACTG-ARL uses a hierarchical DP feature-learning plus conditional-generation pipeline followed by anchored RL to raise MAUVE scores by 20% and improve instruction following under privacy constraints.
-
NashPG: A Policy Gradient Method with Iteratively Refined Regularization for Finding Nash Equilibria
NashPG is a policy-gradient method with iteratively refined regularization that guarantees monotonic convergence to Nash equilibria in two-player zero-sum extensive-form games and scales to large benchmarks.
-
SSL4RL: Revisiting Self-supervised Learning as Intrinsic Reward for Visual-Language Reasoning
SSL4RL reformulates self-supervised learning objectives into dense, verifiable reward signals for RL-based fine-tuning of vision-language models, yielding performance gains on reasoning benchmarks.
-
Vision-LLMs for Spatiotemporal Traffic Forecasting
ST-Vision-LLM reframes spatiotemporal traffic forecasting as vision-language fusion, using visual encoders on traffic grids and efficient numerical tokenization to achieve 15.6% better long-term accuracy and 30% gains in few-shot cross-domain settings.
-
Learning to Trust: Dynamic Utilization of Retrieval-Augmented Generation for E-commerce Search Relevance
DyKnow-RAG uses Group Relative Policy Optimization with dual-group rollouts and posterior-driven advantage scaling to optimize context utilization in RAG for e-commerce relevance, showing offline gains and production lifts when deployed at Taobao.
-
ViSurf: Visual Supervised-and-Reinforcement Fine-Tuning for Large Vision-and-Language Models
ViSurf unifies SFT and RLVR for LVLMs in one training stage by injecting ground-truth labels into rollouts and applying novel reward controls, outperforming standalone and two-stage baselines on diverse benchmarks.
-
On the optimization dynamics of RLVR: Gradient gap and step size thresholds
The paper defines a Gradient Gap for RLVR policy gradients and proves a sharp step-size threshold below which training converges and above which it collapses, with predictions for length and success-rate scaling validated in simulations and on Qwen2.5-Math-7B.
-
Red-Bandit: Test-Time Adaptation for LLM Red-Teaming via Bandit-Guided LoRA Experts
Red-Bandit adapts online to LLM failure modes by dynamically selecting among RL-trained LoRA attack-style experts via a bandit policy, reporting SOTA ASR@10 on AdvBench with lower-perplexity prompts.
-
When Importance Sampling Misallocates Credit: Asymmetric Ratios for Outcome-Supervised RL
The paper identifies that importance sampling ratios in outcome-supervised RL misallocate credit by creating unbalanced token updates, and introduces ASPO to correct the asymmetry for positive-advantage tokens.
-
A KL-regularization Framework for Learning to Plan with Adaptive Priors
PO-MPC unifies prior MPPI-based RL approaches under a single KL-regularized framework that uses the planner distribution as a prior, with new variations yielding performance gains in experiments.
-
Use the Online Network If You Can: Towards Fast and Stable Reinforcement Learning
MINTO sets bootstrapped targets to the minimum of online and target network estimates, yielding faster stable value learning across online/offline RL and discrete/continuous actions.
-
Thinking Sparks!: Emergent Attention Heads in Reasoning Models During Post Training
Post-training on reasoning tasks sparks the emergence of specialized attention heads that enable structured computation, with SFT adding stable heads while GRPO uses dynamic activation and pruning tied to reward signals, and controllable think models relying on compensatory heads instead of specific
-
World-Env: Leveraging World Model as a Virtual Environment for VLA Post-Training
World-Env replaces physical robot interactions with a world model-based virtual environment and VLM-guided rewards to enable efficient RL post-training for VLA models, showing gains with only five demonstrations per task.
-
Structured In-context Environment Scaling for Large Language Model Reasoning
SIE framework automatically constructs scalable, verifiable reasoning environments from structured data, improving in-domain performance and enabling generalization to out-of-domain math and logic tasks.
-
Mitigating Visual Context Degradation in Large Multimodal Models: A Training-Free Decoupled Agentic Framework
DRP decouples reasoning from perception in LMMs by using an LLM reasoner to query an LMM observer for visual details as needed, reducing visual grounding loss.
-
Multiplayer Nash Preference Optimization
MNPO extends NLHF to multiplayer Nash games, inheriting equilibrium guarantees while showing empirical gains on instruction-following benchmarks under diverse preferences.
-
Reinforcement Learning with Discrete Diffusion Policies for Combinatorial Action Spaces
A method trains discrete diffusion policies for combinatorial RL by matching to a PMD-regularized target distribution, reporting SOTA performance and sample efficiency on DNA generation, macro-action, and multi-agent benchmarks.
-
MoveFM-R: Advancing Mobility Foundation Models via Language-driven Semantic Reasoning
MoveFM-R is a framework that bridges mobility foundation models and LLMs using semantically enhanced location encoding, progressive curriculum alignment, and interactive self-reflection to generate plausible trajectories from language inputs.
-
GIFT: Guided Importance-Aware Fine-Tuning for Diffusion Language Models
GIFT weights tokens by entropy during fine-tuning of diffusion language models and reports better performance than standard SFT on reasoning benchmarks across multiple settings.
-
CE-GPPO: Coordinating Entropy via Gradient-Preserving Clipping Policy Optimization in Reinforcement Learning
CE-GPPO preserves bounded gradients from clipped tokens in PPO to regulate entropy evolution and improve performance on mathematical reasoning benchmarks.
-
High-Precision and High-Efficiency Trajectory Tracking for Excavators Based on Closed-Loop Dynamics
EfficientTrack integrates model-based learning and closed-loop dynamics to minimize tracking errors in excavator trajectory control with high efficiency and precision, outperforming prior learning-based methods in simulation and real-world tests.
-
Compute as Teacher: Turning Inference Compute Into Reference-Free Supervision
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.
-
Dissecting Discrete Soft Actor-Critic: Limitations and Principled Alternatives
Shows entropy coupling limits DSAC on discrete tasks and introduces a generalized actor-critic framework with m-step critics and novel entropy-regularized objectives that perform robustly on Atari.
-
SimpleVLA-RL: Scaling VLA Training via Reinforcement Learning
SimpleVLA-RL applies tailored reinforcement learning to VLA models, reaching SoTA on LIBERO, outperforming π₀ on RoboTwin, and surpassing SFT in real-world tasks while reducing data needs and identifying a 'pushcut' phenomenon.
-
Beyond Correctness: Harmonizing Process and Outcome Rewards through RL Training
PROF curates RL training data via PRM-ORM consistency to improve both final-answer accuracy and intermediate reasoning quality while reducing reliance on strong process reward models.
-
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.
-
Scalable Option Learning in High-Throughput Environments
SOL is a new hierarchical RL algorithm that reaches 35x higher throughput and outperforms flat agents when trained on 30 billion frames in NetHack while showing positive scaling.
-
InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency
InternVL3.5 advances open-source multimodal models with Cascade RL for +16% reasoning gains and ViR for 4x inference speedup, with the 241B model reaching SOTA among open-source MLLMs on multimodal, reasoning, and agentic tasks.
-
Beyond Memorization: Extending Reasoning Depth with Recurrence, Memory and Test-Time Compute Scaling
In a cellular automata rule-inference task designed to block memorization, neural models achieve high next-step accuracy but accuracy falls sharply with longer reasoning chains; depth, recurrence, memory, and test-time compute extend the reachable depth but do not remove the bound.
-
Reinforcement Learning-based Control via Y-wise Affine Neural Networks (YANNs)
YANN-RL initializes RL actor and critic networks with explicit multi-parametric linear MPC solutions via YANNs to start from linear optimal control performance and then learn nonlinear policies through online interaction.
-
Robust Remote Reinforcement Learning over Unreliable Communication Channels using Homomorphic State Encoding
HR3L enables robust remote RL training over unreliable channels via homomorphic state encoding without gradient exchange, outperforming prior methods in sample efficiency and adapting to packet loss, delays, and bandwidth limits.
-
RL-PLUS: Countering Capability Boundary Collapse of LLMs in Reinforcement Learning with Hybrid-policy Optimization
RL-PLUS is a hybrid RL approach for LLMs that combines internal exploitation with external data via importance sampling and exploration advantages to prevent capability boundary collapse and achieve gains on math and OOD reasoning benchmarks.
-
TARS: MinMax Token-Adaptive Preference Strategy for Hallucination Reduction in MLLMs
TARS uses token-adaptive min-max preference optimization and FFT-based spectral regularization to cut hallucination rates in MLLMs from 26.4% to 13.2% with only 4.8k samples, outperforming standard DPO and larger data-augmented baselines.
-
Stabilizing Knowledge, Promoting Reasoning: Dual-Token Constraints for RLVR
Archer introduces response-level entropy normalization and differentiated clipping/KL regularization in RLVR to encourage exploration on reasoning tokens while stabilizing knowledge tokens, yielding gains in pass@1 and pass@K on reasoning benchmarks.
-
AnyPos: Automated Task-Agnostic Actions for Bimanual Manipulation
AnyPos automates task-agnostic action collection and inverse-dynamics modeling with arm/end-effector decoupling plus a direction-aware decoder, delivering 51% higher test accuracy and 30-40% better success rates on bimanual tasks.
-
Mitigating Object Hallucinations via Sentence-Level Early Intervention
SENTINEL reduces MLLM object hallucinations by over 90% via sentence-level early intervention with detector-bootstrapped preference data and C-DPO loss, outperforming prior SOTA on hallucination and capability benchmarks.
-
Perception-Aware Policy Optimization for Multimodal Reasoning
PAPO integrates perception-aware supervision via a KL-based loss into RLVR methods like GRPO, yielding 4.4-17.5% gains on multimodal benchmarks and 30.5% fewer perception errors, with larger gains on vision-heavy tasks.
-
MemAgent: Reshaping Long-Context LLM with Multi-Conv RL-based Memory Agent
MemAgent uses multi-conversation RL to train a memory agent that reads text in segments and overwrites memory, extrapolating from 8K training to 3.5M token QA with under 5% loss and 95%+ on 512K RULER.
-
Listener-Rewarded Thinking in VLMs for Image Preferences
Listener-augmented GRPO uses an independent frozen VLM to provide dense confidence scores on reasoning traces, yielding 67.4% accuracy on ImageReward, up to +6% OOD gains on 1.2M-vote human data, and fewer reasoning contradictions.
-
A Survey of Continual Reinforcement Learning
The paper surveys CRL literature, proposes a taxonomy of methods into four categories based on knowledge storage and transfer, reviews metrics and benchmarks, and outlines challenges and future research directions.
-
LongWriter-Zero: Mastering Ultra-Long Text Generation via Reinforcement Learning
LongWriter-Zero applies RL from a base model with specialized rewards for length, quality, and structure to outperform SFT baselines and larger models on long-writing benchmarks.