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%).
1883
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.
DeFAb is a large-scale, formally verifiable benchmark for defeasible abduction derived from 18 knowledge bases, demonstrating that frontier LLMs achieve 7.8-65% accuracy versus 100% for a rule-based solver with polynomial-time checks.
Machine learning discovers a tube-seeding strategy for IBP reduction of Feynman integrals that scales linearly with numerator power, demonstrated on rank-20 2-loop 5-point integrals.
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.
citing papers explorer
-
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.
-
MEM1: Learning to Synergize Memory and Reasoning for Efficient Long-Horizon Agents
MEM1 uses end-to-end RL to learn constant-memory agents that update a shared state for memory and reasoning, delivering 3.5x better performance and 3.7x lower memory use than larger baselines on long-horizon QA and shopping tasks.
-
MiniMax-M1: Scaling Test-Time Compute Efficiently with Lightning Attention
MiniMax-M1 is a 456B parameter hybrid-attention MoE model trained with CISPO RL that achieves performance comparable or superior to DeepSeek-R1 and Qwen3-235B on reasoning and software engineering tasks while training in three weeks on 512 GPUs.
-
Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoning
High-entropy minority tokens drive RLVR gains, so restricting gradients to the top 20% maintains or improves performance over full updates on Qwen3 models, especially larger ones.
-
AReaL: A Large-Scale Asynchronous Reinforcement Learning System for Language Reasoning
AReaL decouples generation and training in LLM reinforcement learning to achieve up to 2.77x speedup with matched or better performance on math and code benchmarks.
-
LoVeC: Reinforcement Learning for Better Verbalized Confidence in Long-Form Generations
LoVeC uses RL to train LLMs to output verbalized numerical confidence scores for statements in long-form text, achieving better calibration than self-consistency baselines on QA datasets while being 20x faster.
-
MUSEG: Reinforcing Video Temporal Understanding via Timestamp-Aware Multi-Segment Grounding
MUSEG applies timestamp-aware multi-segment grounding with a phased-reward RL recipe to boost temporal grounding and time-sensitive video QA performance in MLLMs.
-
Real-World Doctor Agent with Proactive Consultation through Multi-Agent Reinforcement Learning
DoctorAgent-RL trains a Qwen2.5-7B doctor agent via multi-agent RL on the new MTMedDialog dataset to conduct dynamic, question-driven consultations, reaching 70% exact diagnostic match in real-patient trials.
-
RANDPOL: Parameter-Efficient End-to-End Quadruped Locomotion via Randomized Policy Learning
RANDPOL achieves effective quadruped locomotion by training only the final linear readout of a randomly initialized and fixed neural network policy, matching PPO results with reduced parameters and enabling zero-shot sim-to-real transfer on Unitree Go2.
-
DreamPolicy: A Unified World-model Policy for Scalable Humanoid Locomotion
DreamPolicy integrates an autoregressive diffusion world model with policy learning to produce a single scalable policy that generalizes to unseen composite terrains for humanoid locomotion.
-
VLA-RL: Towards Masterful and General Robotic Manipulation with Scalable Reinforcement Learning
VLA-RL applies online RL to pretrained VLAs, yielding a 4.5% gain over strong baselines on 40 LIBERO manipulation tasks and matching commercial models like π₀-FAST.
-
Interactive Post-Training for Vision-Language-Action Models
RIPT-VLA applies RL with dynamic rollout sampling and leave-one-out advantage estimation to fine-tune VLA models, achieving up to 97.5% success rates and recovering from 4% to 97% success with one demonstration in 15 iterations.
-
LLaDA-V: Large Language Diffusion Models with Visual Instruction Tuning
LLaDA-V is a diffusion-based multimodal large language model that reaches competitive or state-of-the-art results on visual instruction tasks while using a non-autoregressive architecture.
-
MMaDA: Multimodal Large Diffusion Language Models
MMaDA is a unified multimodal diffusion model using mixed chain-of-thought fine-tuning and a new UniGRPO reinforcement learning algorithm that outperforms specialized models in reasoning, understanding, and text-to-image tasks.
-
The Unreasonable Effectiveness of Entropy Minimization in LLM Reasoning
Entropy minimization on self-generated outputs elicits strong reasoning in pretrained LLMs, matching or exceeding supervised RL methods on benchmarks.
-
Unconventional Hexacopters via Evolution and Learning: Performance Gains and New Insights
Evolving hexacopter morphologies together with learnable controllers produces unconventional drones that outperform standard designs on complex tasks while introducing new metrics for evolution-learning interactions.
-
DanceGRPO: Unleashing GRPO on Visual Generation
DanceGRPO applies GRPO to visual generation tasks to achieve stable policy optimization across diffusion models, rectified flows, multiple tasks, and diverse reward models, outperforming prior RL methods.
-
ZeroSearch: Incentivize the Search Capability of LLMs without Searching
ZeroSearch uses supervised fine-tuning to create a simulated retrieval module and curriculum-based RL rollouts that degrade document quality to train LLMs on search capabilities without real search API calls.
-
Learning to Reason under Off-Policy Guidance
LUFFY mixes off-policy reasoning traces into RLVR training via Mixed-Policy GRPO and regularized importance sampling, delivering over 6-point gains on math benchmarks and enabling training of weak models where on-policy RLVR fails.
-
ToolRL: Reward is All Tool Learning Needs
A principled reward design for tool selection and application in RL-trained LLMs delivers 17% gains over base models and 15% over SFT across benchmarks.
-
ReTool: Reinforcement Learning for Strategic Tool Use in LLMs
ReTool uses outcome-driven RL to train 32B LLMs to dynamically use code tools during reasoning, reaching 72.5% accuracy on AIME and surpassing o1-preview.
-
VLM-R1: A Stable and Generalizable R1-style Large Vision-Language Model
VLM-R1 applies R1-style RL using rule-based rewards on visual tasks with clear ground truth to achieve competitive performance and superior generalization over SFT in vision-language models.
-
ShadowCoT: Cognitive Hijacking for Stealthy Reasoning Backdoors in LLMs
ShadowCoT introduces a reasoning-level backdoor attack on LLMs achieving 94.4% attack success rate and 88.4% hijacking success rate with 0.15% parameter updates via internal state conditioning and reasoning chain pollution.
-
VAPO: Efficient and Reliable Reinforcement Learning for Advanced Reasoning Tasks
VAPO achieves 60.4 on AIME 2024 with Qwen 32B, outperforming prior methods by over 10 points through targeted fixes for value bias, sequence length variation, and sparse rewards.
-
Joint Optimization of Handoff and Video Rate in LEO Satellite Networks
The paper introduces MPC and RL algorithms for joint satellite handoff and video bitrate selection to optimize QoE in LEO networks, validated in trace-driven simulations and testbed experiments.
-
Rollout Then Optimize: A One-Step Newton Refinement of Learned Policies for Nonlinear Model Predictive Control
A learned policy rollout is refined by one Newton step via Riccati recursion, yielding quadratic reduction in suboptimality for nonlinear MPC, shown on quadcopter trajectory tracking.
-
Open-Reasoner-Zero: An Open Source Approach to Scaling Up Reinforcement Learning on the Base Model
A simple PPO-based RL training pipeline on base models scales reasoning performance and response length, outperforming prior work on math and science benchmarks with one-tenth the training steps.
-
UI-R1: Enhancing Efficient Action Prediction of GUI Agents by Reinforcement Learning
UI-R1 shows rule-based RL with GRPO on 136 GUI tasks improves a 3B MLLM's action prediction accuracy by 6-22% over its base model and matches larger SFT-trained models.
-
ReSearch: Learning to Reason with Search for LLMs via Reinforcement Learning
ReSearch trains LLMs via RL to integrate search operations into reasoning steps, achieving strong generalization across benchmarks and eliciting reflection and self-correction without supervised reasoning data.
-
DAPO: An Open-Source LLM Reinforcement Learning System at Scale
DAPO introduces decoupled clipping and dynamic sampling for LLM RL, achieving 50 on AIME 2024 with Qwen2.5-32B while fully open-sourcing code, data, and the verl-based training system.
-
Characterizing higher-order representations through generative diffusion models explains human decoded neurofeedback performance
NERD uses RL-trained diffusion models on fMRI data to model higher-order uncertainty representations, outperforming controls and linking individual differences to neurofeedback success.
-
LMM-R1: Empowering 3B LMMs with Strong Reasoning Abilities Through Two-Stage Rule-Based RL
A two-stage RL framework first boosts text reasoning in 3B LMMs then adapts it to multimodal inputs, producing modest benchmark gains of 4.5-4.8%.
-
Visual-RFT: Visual Reinforcement Fine-Tuning
Visual-RFT applies reinforcement learning with verifiable perception rewards to improve large vision-language models on fine-grained classification, few-shot detection, and grounding tasks.
-
Supervising the search process produces reliable and generalizable information-seeking agents
Process supervision via RAG-Gym produces more reliable and generalizable search agents, with gains driven by higher-quality queries on out-of-domain multi-hop tasks.
-
Step-Audio: Unified Understanding and Generation in Intelligent Speech Interaction
Step-Audio introduces a 130B-parameter unified speech-text model with open-sourced components for understanding, generation, affordable voice cloning, and dynamic control, claiming SOTA human evaluation results on a new benchmark.
-
Process Reinforcement through Implicit Rewards
PRIME enables online process reward model updates in LLM RL using implicit rewards from rollouts and outcome labels, yielding 15.1% average gains on reasoning benchmarks and surpassing a stronger instruct model with 10% of the data.
-
The Differences Between Direct Alignment Algorithms are a Blur
A controlled unification of direct alignment algorithms shows the ranking objective (pairwise vs pointwise) drives alignment quality more than the scalar score optimized.
-
Asynchronous Cooperative Multi-Agent Reinforcement Learning with Limited Communication
AsynCoMARL is a new asynchronous MARL algorithm that matches leading baselines on success and collision rates while using 26% fewer messages via graph transformers on dynamic communication graphs.
-
SFT Memorizes, RL Generalizes: A Comparative Study of Foundation Model Post-training
Reinforcement learning post-training enables generalization to unseen textual rule variants and visual changes in foundation models, while supervised fine-tuning primarily leads to memorization.
-
Dream to Fly: Model-Based Reinforcement Learning for Vision-Based Drone Flight
DreamerV3 enables pixel-to-control policies for drone racing that reach 9 m/s in both simulation and real hardware-in-the-loop tests.
-
Improving Video Generation with Human Feedback
A human preference dataset and VideoReward model enable Flow-DPO and Flow-NRG to produce smoother, better-aligned videos from text prompts in flow-based generators.
-
HuatuoGPT-o1, Towards Medical Complex Reasoning with LLMs
HuatuoGPT-o1 achieves superior medical complex reasoning by using a verifier to curate reasoning trajectories for fine-tuning and then applying RL with verifier-based rewards.
-
RoboMD: Uncovering Robot Vulnerabilities through Semantic Potential Fields
A deep RL vulnerability-prediction policy trained in semantic embedding space finds up to 23% more unique robot manipulation failures than vision-language baselines and enables more efficient fine-tuning.
-
Enhancing the Reasoning Ability of Multimodal Large Language Models via Mixed Preference Optimization
Mixed Preference Optimization with the MMPR dataset boosts multimodal CoT reasoning, lifting InternVL2-8B to 67.0 accuracy on MathVista (+8.7 points) and matching the 76B model.
-
The Hive Mind is a Single Reinforcement Learning Agent
Bee hive mind from weighted voter imitation equals a single RL agent using a new multi-armed bandit rule called Maynard-Cross Learning.
-
HybridFlow: A Flexible and Efficient RLHF Framework
HybridFlow combines single- and multi-controller paradigms with a 3D-HybridEngine to deliver 1.53x to 20.57x higher throughput for various RLHF algorithms compared to prior systems.
-
Training Language Models to Self-Correct via Reinforcement Learning
SCoRe uses multi-turn online RL with regularization on self-generated traces to improve LLM self-correction, achieving 15.6% and 9.1% gains on MATH and HumanEval for Gemini models.
-
Diffusion Policy Policy Optimization
DPPO fine-tunes diffusion policies via policy gradients and outperforms prior RL approaches for diffusion policies and PG-tuned alternatives on robot benchmarks while enabling stable training and hardware deployment.
-
Inference Scaling Laws: An Empirical Analysis of Compute-Optimal Inference for Problem-Solving with Language Models
Empirical analysis shows scaling inference compute via strategies like tree search can be more efficient than scaling model parameters, with 7B models plus novel search outperforming 34B models.
-
Learning Multi-Modal Whole-Body Control for Real-World Humanoid Robots
A single learned controller called MHC enables real humanoid robots to execute diverse whole-body behaviors from multi-modal inputs via masked target trajectories.