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%).
1603
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
-
Pref-CTRL: Preference Driven LLM Alignment using Representation Editing
Pref-CTRL trains a multi-objective value function on preferences to guide representation editing for LLM alignment, outperforming RE-Control on benchmarks with better out-of-domain generalization.
-
V-GRPO: Online Reinforcement Learning for Denoising Generative Models Is Easier than You Think
V-GRPO makes ELBO surrogates stable and efficient for online RL alignment of denoising models, delivering SOTA text-to-image performance with 2-3x speedups over MixGRPO and DiffusionNFT.
-
Process Supervision of Confidence Margin for Calibrated LLM Reasoning
RLCM trains LLMs with a margin-enhanced process reward that widens the gap between correct and incorrect reasoning steps, improving calibration on math, code, logic, and science tasks without hurting accuracy.
-
Reducing Detail Hallucinations in Long-Context Regulatory Understanding via Targeted Preference Optimization
DetailDPO cuts detail-level hallucination errors in LLMs on long regulatory documents by 42-61% using targeted contrastive pairs on a new 13,000-pair benchmark.
-
QDTraj: Exploration of Diverse Trajectory Primitives for Articulated Objects Robotic Manipulation
QDTraj uses Quality-Diversity algorithms with sparse rewards to produce at least five times more diverse high-performing trajectories for articulated object manipulation than compared methods, validated across 30 objects with hundreds of trajectories per task.
-
Qubit-Scalable CVRP via Lagrangian Knapsack Decomposition and Noise-Aware Quantum Execution
A hybrid quantum framework decomposes CVRP into bounded-width knapsack subproblems, trains a reinforcement learning controller for Lagrangian multipliers, and uses a contextual bandit to adapt quantum hardware execution, yielding improved routing quality on standard test instances.
-
A Hybrid Reinforcement and Self-Supervised Learning Aided Benders Decomposition Algorithm
A hybrid RL and self-supervised learning method accelerates generalized Benders decomposition by 57.5% on a MINLP case study while recovering optimal solutions.
-
Insect-inspired modular architectures as inductive biases for reinforcement learning
Modular insect-inspired RL architecture with arbitration outperforms centralized GRU and MLP baselines on a multi-objective navigation task, showing higher returns, lower value loss, and selective module use.
-
Alignment has a Fantasia Problem
AI alignment must move beyond assuming users have fully formed goals and instead provide active cognitive support to help form and refine intent over time.
-
Generative Learning Enhanced Intelligent Resource Management for Cell-Free Delay Deterministic Communications
The proposed pretraining framework for safe DRL in CF-MIMO resource management doubles initial energy efficiency, achieves 4.7% higher final EE, maintains 1% delay violation rate, and cuts exploration steps by 50% compared to non-pretrained baselines while matching diffusion model performance at 14x
-
X2-N: A Transformable Wheel-legged Humanoid Robot with Dual-mode Locomotion and Manipulation
X2-N is a transformable wheel-legged humanoid robot with a reinforcement learning whole-body controller that enables dual-mode locomotion and manipulation across varied terrains.
-
GeoMind: An Agentic Workflow for Lithology Classification with Reasoned Tool Invocation
GeoMind applies an agentic workflow with tool-augmented modules and process supervision to outperform static models on lithology classification from well logs while producing traceable decisions.
-
Tempered Sequential Monte Carlo for Trajectory and Policy Optimization with Differentiable Dynamics
Tempered sequential Monte Carlo samples from a Boltzmann-tilted distribution over controllers to optimize trajectories and policies under differentiable dynamics.
-
Conjecture and Inquiry: Quantifying Software Performance Requirements via Interactive Retrieval-Augmented Preference Elicitation
IRAP quantifies ambiguous performance requirements into mathematical functions via interactive retrieval-augmented preference elicitation and outperforms ten prior methods on four real-world datasets with up to 40x gains in five interaction rounds.
-
ReaGeo: Reasoning-Enhanced End-to-End Geocoding with LLMs
ReaGeo is an end-to-end LLM framework for geocoding that uses geohash text generation, Chain-of-Thought spatial reasoning, and distance-based RL to accurately predict points and regions from explicit and vague queries.
-
Navigating the Clutter: Waypoint-Based Bi-Level Planning for Multi-Robot Systems
Waypoint-based bi-level planning with curriculum RLVR improves multi-robot task success rates in dense-obstacle benchmarks over motion-agnostic and VLA baselines.
-
SSL-R1: Self-Supervised Visual Reinforcement Post-Training for Multimodal Large Language Models
SSL-R1 reformulates visual SSL tasks into verifiable puzzles to supply rewards for RL post-training of MLLMs, yielding gains on multimodal benchmarks without external supervision.
-
GRPO-VPS: Enhancing Group Relative Policy Optimization with Verifiable Process Supervision for Effective Reasoning
GRPO-VPS improves GRPO by using segment-wise conditional probabilities of the correct answer to supply process-level feedback, yielding up to 2.6-point accuracy gains and 13.7% shorter reasoning on math tasks.
-
WebGen-R1: Incentivizing Large Language Models to Generate Functional and Aesthetic Websites with Reinforcement Learning
WebGen-R1 uses end-to-end RL with scaffold-driven generation and cascaded rewards for structure, function, and aesthetics to transform a 7B model into a generator of deployable multi-page websites that rivals much larger models.
-
Building a Precise Video Language with Human-AI Oversight
CHAI framework pairs AI pre-captions with expert human critiques to produce precise video descriptions, enabling open models to outperform closed ones like Gemini-3.1-Pro and improve fine-grained control in video generation models.
-
ETac: A Lightweight and Efficient Tactile Simulation Framework for Learning Dexterous Manipulation
ETac is a data-driven tactile simulation framework that matches FEM deformation accuracy at high speed, supporting 4096 parallel environments at 869 FPS and yielding 84.45% success in blind grasping across four object types.
-
Hybrid Policy Distillation for LLMs
Hybrid Policy Distillation unifies existing knowledge distillation methods for LLMs into a reweighted log-likelihood objective and introduces a hybrid forward-reverse KL approach with mixed data sampling to improve stability, efficiency, and performance.
-
Scaling Self-Play with Self-Guidance
SGS adds self-guidance to LLM self-play for Lean4 theorem proving, surpassing RL baselines and enabling a 7B model to outperform a 671B model after 200 rounds.
-
A Delta-Aware Orchestration Framework for Scalable Multi-Agent Edge Computing
DAOEF integrates delta-aware caching, action pruning, and hardware matching to deliver 1.45x gains and sub-linear scaling up to 250 agents in multi-agent edge computing.
-
Bootstrapping Post-training Signals for Open-ended Tasks via Rubric-based Self-play on Pre-training Text
POP bootstraps post-training signals for open-ended LLM tasks by synthesizing rubrics during self-play on pretraining corpus, yielding performance gains on Qwen-2.5-7B across healthcare QA, creative writing, and instruction following.
-
Safe Continual Reinforcement Learning in Non-stationary Environments
Safe continual RL methods face a fundamental tension between enforcing safety constraints and preventing catastrophic forgetting in non-stationary environments, with regularization providing only partial mitigation.
-
Pause or Fabricate? Training Language Models for Grounded Reasoning
GRIL uses stage-specific RL rewards to train LLMs to detect missing premises, pause proactively, and resume grounded reasoning after clarification, yielding up to 45% better premise detection and 30% higher task success on insufficient math datasets.
-
DR-MMSearchAgent: Deepening Reasoning in Multimodal Search Agents
DR-MMSearchAgent derives batch-wide trajectory advantages and uses differentiated Gaussian rewards to prevent premature collapse in multimodal agents, outperforming MMSearch-R1 by 8.4% on FVQA-test.
-
Multi-Gait Learning for Humanoid Robots Using Reinforcement Learning with Selective Adversarial Motion Prior
Selective AMP in RL enables a single policy for five humanoid gaits with faster convergence and better performance on stability tasks without losing dynamic agility.
-
SynAgent: Generalizable Cooperative Humanoid Manipulation via Solo-to-Cooperative Agent Synergy
SynAgent enables generalizable cooperative humanoid manipulation by transferring skills from solo human-object interactions to multi-agent scenarios via interaction-preserving retargeting, single-agent pretraining with multi-agent PPO, and a conditional VAE generative policy.
-
S2H-DPO: Hardness-Aware Preference Optimization for Vision-Language Models
S2H-DPO generates hierarchical prompt-driven preference pairs to improve multi-image reasoning in VLMs while keeping single-image performance intact.
-
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.
-
Towards Disentangled Preference Optimization Dynamics: Suppress the Loser, Preserve the Winner
A unified incentive-score decomposition of preference optimization reveals the disentanglement band condition and reward calibration method that enables suppressing losers while preserving winners in LLM training.
-
QuantumQA: Enhancing Scientific Reasoning via Physics-Consistent Dataset and Verification-Aware Reinforcement Learning
QuantumQA dataset and verification-aware RL with adaptive reward fusion enable an 8B LLM to achieve performance competitive with proprietary models on quantum mechanics tasks.
-
Mitigating Multimodal Hallucination via Phase-wise Self-reward
PSRD mitigates visual hallucinations in LVLMs via phase-wise self-reward decoding, cutting rates by 50% on LLaVA-1.5-7B and outperforming prior methods on five benchmarks.
-
HEALing Entropy Collapse: Enhancing Exploration in Few-Shot RLVR via Hybrid-Domain Entropy Dynamics Alignment
HEAL mitigates entropy collapse in few-shot RLVR by selectively adding general-domain data and aligning trajectory-level entropy dynamics, matching full-shot performance with 32 target samples.
-
Re$^2$MoGen: Open-Vocabulary Motion Generation via LLM Reasoning and Physics-Aware Refinement
Re²MoGen generates open-vocabulary motions via MCTS-enhanced LLM keyframe planning, pose-prior optimization with dynamic temporal matching fine-tuning, and physics-aware RL post-training, claiming SOTA performance.
-
Beyond Static Snapshots: A Grounded Evaluation Framework for Language Models at the Agentic Frontier
ISOPro replaces learned reward models with deterministic verifiers in a continuous evaluation setup for LLMs, delivering larger average capability gains than GRPO-LoRA across small models in scheduling and MBPP domains while characterizing a buffer-skew failure mode.
-
Cat-DPO: Category-Adaptive Safety Alignment
Cat-DPO applies per-category adaptive safety margins during direct preference optimization to reduce variance in safety across harm categories.
-
Learning to Control Summaries with Score Ranking
A score-ranking loss enables controllable summarization by aligning outputs to evaluation scores, matching SOTA performance with dimension-specific control on LLaMA, Qwen, and Mistral.
-
Abstain-R1: Calibrated Abstention and Post-Refusal Clarification via Verifiable RL
A 3B model trained via clarification-aware RLVR improves abstention and post-refusal clarification on unanswerable queries while matching larger models like DeepSeek-R1 on benchmarks.
-
SPS: Steering Probability Squeezing for Better Exploration in Reinforcement Learning for Large Language Models
SPS interleaves RL and IRL to counteract probability squeezing in LLM reasoning trajectories, improving Pass@k on five benchmarks while identifying an empirical upper bound on multi-sample performance.
-
Incentivizing Parametric Knowledge via Reinforcement Learning with Verifiable Rewards for Cross-Cultural Entity Translation
EA-RLVR boosts Qwen3-14B entity translation accuracy from 23.66% to 31.87% on 50k unseen entities using 7k samples via RL with verifiable rewards, with transfer gains of +1.35 XCOMET on WMT24++.
-
GRAIL: Autonomous Concept Grounding for Neuro-Symbolic Reinforcement Learning
GRAIL autonomously grounds relational concepts in NeSy-RL by using LLM weak supervision followed by interaction-based refinement, matching or exceeding manually defined concepts on Atari games.
-
AutoOR: Scalably Post-training LLMs to Autoformalize Operations Research Problems
AutoOR uses synthetic data generation and RL post-training with solver feedback to enable 8B LLMs to autoformalize linear, mixed-integer, and non-linear OR problems, matching larger models on benchmarks.
-
SocialGrid: A Benchmark for Planning and Social Reasoning in Embodied Multi-Agent Systems
SocialGrid benchmark shows even top LLMs achieve below 60% in embodied planning and task completion, with deception detection near random chance regardless of model scale.
-
Scattered Hypothesis Generation for Open-Ended Event Forecasting
SCATTER uses RL with a hybrid reward combining validity, intra-group diversity, and inter-group diversity to produce inclusive hypothesis sets for event forecasting and outperforms baselines on OpenForecast and OpenEP.
-
GroupDPO: Memory efficient Group-wise Direct Preference Optimization
GroupDPO decouples group-wise preference optimization during backpropagation to cut peak memory while keeping the same gradients, allowing larger groups and consistent gains over single-pair DPO plus an NLL term on positives.
-
Reward Weighted Classifier-Free Guidance as Policy Improvement in Autoregressive Models
Reward-weighted classifier-free guidance approximates Q-function policy improvement in autoregressive models, enabling test-time reward optimization and faster RL convergence via distillation.
-
Efficient $n$-qubit entangling operations via a superconducting quantum router
A superconducting quantum router enables programmable multi-qubit entangling operations, demonstrated with faster preparation of entangled states and RL-trained 2- and 3-qubit gates like Toffoli and Fredkin.