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%).
1553
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.
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.
GEAR jointly trains VQ tokenizer and AR generator end-to-end via dual hard/soft read-out and representation alignment, achieving up to 10x faster ImageNet gFID convergence than LlamaGen-REPA while generalizing across quantizers and to text-to-image.
citing papers explorer
-
Do Less, Achieve More: Do We Need Every-Step Optimization for RL Fine-tuning of Diffusion Models?
AdaScope adaptively selects optimal RL intervention points during diffusion denoising by monitoring structural and semantic changes, delivering 66% higher performance at 59% lower cost than full-trajectory RL baselines.
-
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.
-
Offline Reinforcement Learning with Universal Horizon Models
Universal horizon models extend geometric horizon models to arbitrary horizons and apply winsorized distributions for stable offline RL value learning, outperforming baselines on 100 OGBench tasks.
-
DualKV: Shared-Prompt Flash Attention for Efficient RL Training with Large Rollouts and Long Contexts
DualKV eliminates redundant prompt replication in RL training attention kernels via fused dual-KV CUDA operations and token repacking, delivering 1.63-3.82x policy-update speedups while remaining mathematically equivalent to standard attention.
-
Beyond Partner Diversity: An Influence-Based Team Steering Framework for Zero-Shot Human-Machine Teaming
IBTS framework uses influence shaping to improve zero-shot human-machine teaming beyond partner diversity alone, with gains shown in Overcooked-AI simulations and a 30-subject human study.
-
Deep Pre-Alignment for VLMs
Deep Pre-Alignment uses a small VLM perceiver instead of ViT to pre-align visual features with LLM text space, yielding 1.9-3.0 point gains on multimodal benchmarks and 32.9% less language forgetting.
-
Second-Order Actor-Critic Methods for Discounted MDPs via Policy Hessian Decomposition
Develops a second-order actor-critic method for discounted MDPs by decomposing the policy Hessian and justifying a locally constant critic approximation via two-timescale separation.
-
Performance-Driven Policy Optimization for Speculative Decoding with Adaptive Windowing
PPOW uses window-level RL with cost-aware speedup and proximity rewards plus adaptive divergence-aware windowing to reach 6.29-6.52 acceptance lengths and 3.39-4.36x speedups in speculative decoding.
-
Critic-Driven Voronoi-Quantization for Distilling Deep RL Policies to Explainable Models
Critic-Driven Voronoi State Partitioning distills deep RL policies into piecewise-linear models by iteratively adding linear subpolicies in high-value-error regions identified by the critic.
-
Selective Safety Steering via Value-Filtered Decoding
Value-filtered decoding steers LLM outputs for safety at decoding time using a value criterion with an explicit bound on false interventions controlled by one threshold hyperparameter.
-
Resolving Action Bottleneck: Agentic Reinforcement Learning Informed by Token-Level Energy
ActFocus resolves the action bottleneck in agentic RL by reweighting token gradients toward action tokens using observed reward variance and an energy-based uncertainty term, outperforming PPO and GRPO by up to 65 percentage points.
-
Reinforcement Learning with Semantic Rewards Enables Low-Resource Language Expansion without Alignment Tax
Reinforcement learning with semantic rewards lets LLMs gain low-resource language skills without the alignment tax that degrades general capabilities in supervised fine-tuning.
-
CrystalReasoner: Reasoning and RL for Property-Conditioned Crystal Structure Generation
CrystalReasoner combines LLM reasoning traces with physical priors and multi-objective RL to generate valid, stable, and property-conditioned crystal structures.
-
Action-Conditioned Risk Gating for Safety-Critical Control under Partial Observability
Action-conditioned near-term risk prediction gates optimistic and conservative value estimates in RL to approximate risk-sensitive POMDP control, yielding better safety-performance tradeoffs with lower runtime than belief planning baselines.
-
Diagnosing Training Inference Mismatch in LLM Reinforcement Learning
Training-inference mismatch in separated rollout and optimization stages of LLM RL can independently cause training collapse.
-
Flow Field Reconstruction with Sensor Placement Policy Learning
A directional GNN combined with constrained PPO jointly improves flow-field reconstruction accuracy and sensor layout selection in realistic fluid dynamics settings.
-
Distribution Corrected Offline Data Distillation for Large Language Models
A distribution-correction framework for offline LLM reasoning distillation improves accuracy on math benchmarks by adaptively aligning teacher supervision with the student's inference-time distribution.
-
History Anchors: How Prior Behavior Steers LLM Decisions Toward Unsafe Actions
A single consistency instruction with harmful prior actions causes aligned frontier LLMs to select unsafe options at 91-98% rates in high-stakes domains, with escalation and inverse scaling by model size.
-
Learning Responsibility-Attributed Adversarial Scenarios for Testing Autonomous Vehicles
CARS integrates responsibility attribution into adversarial scenario generation to produce physically feasible collisions with high attribution rates under regulation-prescribed driver models for autonomous vehicle testing.
-
Rethinking Molecular OOD Generalization via Target-Aware Source Selection
SCOPE-BENCH shows state-of-the-art molecular models suffer up to 8x higher errors under extreme OOD, while POMA reduces mean absolute error by up to 11.2% via target-aware source selection and dual-scale adaptation.
-
CO-MAP: A Reinforcement Learning Approach to the Qubit Allocation Problem
Reinforcement learning policy for qubit mapping reduces SWAP overhead by 65-85% versus standard quantum compilers on MQTBench and Queko benchmark circuits.
-
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.
-
What to Ignore, What to React: Visually Robust RL Fine-Tuning of VLA Models
PAIR-VLA adds invariance and sensitivity objectives over paired visual variants during PPO fine-tuning of VLA models, yielding 9-16% average gains on ManiSkill3 under distractors, textures, poses, viewpoints, and lighting shifts.
-
Learning to See What You Need: Gaze Attention for Multimodal Large Language Models
Gaze Attention groups visual embeddings into selectable regions and dynamically restricts attention to task-relevant ones, matching dense baselines with up to 90% fewer visual KV entries via added context tokens.
-
From Cumulative Constraints to Adaptive Runtime Safety Control for Nonstationary Reinforcement Learning
CPSS projects cumulative safety constraints into time-varying per-state thresholds for online action shielding in nonstationary RL, providing per-state guarantees and cumulative bounds.
-
Revisiting DAgger in the Era of LLM-Agents
DAgger-style training with turn-level policy interpolation raises 4B and 8B LLM agents to 27.3% and 29.8% on SWE-bench Verified, beating several larger published systems.
-
Robust Instruction Compliance in Cooperative Multi-Agent Reinforcement Learning
MAVIC corrects Bellman backups at instruction boundaries by adjusting the incoming objective and restoring continuation value, enabling consistent estimation under stochastic instruction switching in cooperative MARL.
-
Multi-Rollout On-Policy Distillation via Peer Successes and Failures
MOPD improves on-policy distillation by using peer successes and failures from multiple rollouts to construct more informative teacher signals, yielding consistent gains over baselines on reasoning benchmarks.
-
Trust the Batch, On- or Off-Policy: Adaptive Policy Optimization for RL Post-Training
A new RL objective adapts trust-region and off-policy handling automatically via normalized effective sample size of batch policy ratios, matching tuned baselines without new hyperparameters.
-
Discrete Flow Matching for Offline-to-Online Reinforcement Learning
DRIFT enables stable offline-to-online fine-tuning of CTMC policies in discrete RL via advantage-weighted discrete flow matching, path-space regularization, and candidate-set approximation.
-
Delay-Empowered Causal Hierarchical Reinforcement Learning
DECHRL models causal structures and stochastic delay distributions within hierarchical RL and incorporates them into a delay-aware empowerment objective to improve performance under temporal uncertainty.
-
Hybrid-LoRA: Bridging Full Fine-Tuning and Low-Rank Adaptation for Post-Training
Hybrid-LoRA selectively full fine-tunes modules with high sensitivity to low-rank adaptation using a novel score and applies LoRA elsewhere, matching full fine-tuning at 10% budget and outperforming PEFT baselines by up to 5.65%.
-
TMRL: Diffusion Timestep-Modulated Pretraining Enables Exploration for Efficient Policy Finetuning
TMRL bridges behavioral cloning pretraining and RL finetuning via diffusion noise and timestep modulation to enable controlled exploration, improving sample efficiency and enabling real-world robot training in under one hour.
-
When Policy Entropy Constraint Fails: Preserving Diversity in Flow-based RLHF via Perceptual Entropy
Policy entropy remains constant in flow-matching models during RLHF due to fixed noise schedules while perceptual diversity collapses from mode-seeking policy gradients, so perceptual entropy constraints are introduced to preserve diversity and improve quality.
-
Learning What Matters: Adaptive Information-Theoretic Objectives for Robot Exploration
QOED selects identifiable parameter directions via Fisher matrix eigenspace analysis and modifies exploration objectives to approximate ideal information gain under bounded nuisance assumptions, yielding 21-35% performance gains in robotic tasks.
-
Missing Old Logits in Asynchronous Agentic RL: Semantic Mismatch and Repair Methods for Off-Policy Correction
Missing old logits in async agentic RL entangle discrepancy and staleness terms in PPO off-policy correction; exact acquisition methods and revised PPO-EWMA restore decoupled updates with reported gains in speed and performance.
-
SAGE: A Self-Evolving Agentic Graph-Memory Engine for Structure-Aware Associative Memory
SAGE is a self-evolving agentic graph-memory engine that dynamically constructs and refines structured memory graphs via writer-reader feedback, yielding performance gains on multi-hop QA, open-domain retrieval, and long-term agent benchmarks.
-
Holder Policy Optimisation
HölderPO unifies token-level aggregation in GRPO via the Hölder mean with a tunable p parameter and annealing schedule, delivering 54.9% average accuracy on math benchmarks and 93.8% success on ALFWorld.
-
SAGE: Scalable Automated Robustness Augmentation for LLM Knowledge Evaluation
SAGE trains a rubric-based verifier and an RL-optimized generator on seed human data to scalably augment LLM knowledge benchmarks, matching human-annotated quality on HellaSwag at lower cost and generalizing to MMLU.
-
Stochastic Minimum-Cost Reach-Avoid Reinforcement Learning
Introduces RAPCs and a contraction Bellman operator for cost-optimal policies that satisfy probabilistic reach-avoid specifications in stochastic MDPs, with almost-sure convergence to local optima.
-
On-Policy Self-Evolution via Failure Trajectories for Agentic Safety Alignment
FATE lets LLM agents self-evolve safer behaviors by generating and filtering repairs from their own failure trajectories using verifiers and Pareto optimization.
-
GRAFT: Graph-Tokenized LLMs for Tool Planning
GRAFT internalizes tool dependency graphs via dedicated special tokens in LLMs and applies on-policy context distillation to achieve higher exact sequence matching and dependency legality than prior external-graph methods.
-
Augmented Lagrangian Method for Last-Iterate Convergence for Constrained MDPs
An inexact augmented Lagrangian method with projected Q-ascent yields global last-iterate convergence guarantees for constrained MDP policy optimization, extending from tabular to log-linear and non-linear policies.
-
UNIPO: Unified Interactive Visual Explanation for RL Fine-Tuning Policy Optimization
UNIPO is the first unified interactive visualization tool exposing token-level training dynamics of RL fine-tuning algorithms for LLMs through high-level overviews, step inspectors, and side-by-side comparisons.
-
Taming Extreme Tokens: Covariance-Aware GRPO with Gaussian-Kernel Advantage Reweighting
Covariance-weighted GRPO with Gaussian-kernel reweighting tames extreme tokens to stabilize training and boost reasoning performance over standard GRPO.
-
Decaf: Improving Neural Decompilation with Automatic Feedback and Search
Decaf uses compiler feedback and search to improve neural decompilation, boosting semantic success rate from 26.0% to 83.9% on ExeBench Real -O2 split.
-
Understanding and Preventing Entropy Collapse in RLVR with On-Policy Entropy Flow Optimization
OPEFO prevents entropy collapse in RLVR by rescaling token updates according to their entropy change contributions, yielding more stable optimization and better results on math benchmarks.
-
Robust Multi-Agent Path Finding under Observation Attacks: A Principled Adversarial-Plus-Smoothing Training Recipe
Adv-PPO and Adv-PPO+MACER raise worst-case success in decentralized MAPF from 2.5% to 77.5% under observation attacks on 8x8 POGEMA maps with four agents, at under 1% clean-performance cost.
-
Adaptive Teacher Exposure for Self-Distillation in LLM Reasoning
ATESD introduces a Beta-policy controller that adapts teacher exposure ratio during LLM self-distillation training and reports gains over fixed-exposure baselines on math benchmarks.
-
Freeze Deep, Train Shallow: Interpretable Layer Allocation for Continued Pre-Training
LayerTracer analysis identifies deep LLM layers as stable task-critical regions, leading to a shallow-train deep-freeze strategy that outperforms full fine-tuning on C-Eval and CMMLU.