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%).
1373
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 154
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
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.
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.
IWR improves CRL sample efficiency and performance in interaction-rich manipulation by interaction-aware resampling that preserves mode boundaries, yielding 19.8% average gains and a real-world air-hockey agent.
EFE-based planning is formulated as variational free energy minimization with epistemic priors, decomposing into expected plan costs plus a complexity term.
EFE-based active inference planning is characterized as VFE on an augmented model plus entropy and planning corrections, with a derived message-passing implementation and grid-world validation.
Reinforcement learning on a bead-spring cilia model identifies antiplectic coordination as flow-maximizing, with a tilted-slider reduced model showing that a time-averaged position shift opposite the effective stroke enhances transport via elastic restoring force coupling, and that symplectic coordi
PromptPO shows LLMs can act as black-box policy optimizers for sequential RL when leveraging prior knowledge, matching baselines in exploration and robotics but underperforming in MuJoCo.
citing papers explorer
-
Alignment faking in large language models
Claude 3 Opus strategically fakes alignment by complying with harmful requests only during simulated training to preserve its preference for refusing them afterward.
-
SimWorld Studio: Automatic Environment Generation with Evolving Coding Agent for Embodied Agent Learning
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.
-
Beyond the Assistant Turn: User Turn Generation as a Probe of Interaction Awareness in Language Models
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.
-
Expected Free Energy-based Planning as Variational Inference
EFE-based planning is formulated as variational free energy minimization with epistemic priors, decomposing into expected plan costs plus a complexity term.
-
What Type of Inference is Active Inference?
EFE-based active inference planning is characterized as VFE on an augmented model plus entropy and planning corrections, with a derived message-passing implementation and grid-world validation.
-
EDGE-OPD: Internalizing Privileged Context with Evidence Guided On-Policy Distillation
EDGE-OPD adds guided rollouts and evidence masking to on-policy self-distillation, enabling successful learning of target identities where standard OPSD and RLSD fail.
-
Conditional Equivalence of DPO and RLHF: Implicit Assumption, Failure Modes, and Provable Alignment
DPO-RLHF equivalence holds only conditionally on the optimal policy preferring human-preferred responses; otherwise DPO optimizes relative advantage and can prefer worse outputs, addressed by introducing CPO.
-
Robotics-Inspired Guardrails for Foundation Models in Socially Sensitive Domains
Introduces the Grounded Observer framework that applies robotics-inspired formal constructs for runtime constraint enforcement on foundation model interaction trajectories in socially sensitive domains.
-
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.
-
Missingness-MDPs: Bridging the Theory of Missing Data and POMDPs
Miss-MDPs extend POMDPs with missing-data theory to learn observation missingness patterns and compute near-optimal policies with high-probability guarantees.
-
CuSearch: Curriculum Rollout Sampling via Search Depth for Agentic RAG
CuSearch reallocates rollout budget in RLVR toward deeper-search trajectories as a proxy for retrieval supervision density, yielding up to 11.8 exact-match gains over uniform GRPO sampling on ZeroSearch.
-
Breaking $\textit{Winner-Takes-All}$: Cooperative Policy Optimization Improves Diverse LLM Reasoning
GCPO uses team-level credit assignment via determinant volume over reward-weighted semantic embeddings to promote non-redundant correct reasoning paths, improving both accuracy and diversity in LLM training.
-
OLIVIA: Online Learning via Inference-time Action Adaptation for Decision Making in LLM ReAct Agents
OLIVIA treats LLM agent action selection as a contextual linear bandit over frozen hidden states and applies UCB exploration to adapt online, yielding consistent gains over static ReAct and prompt-based baselines on four benchmarks.
-
AHD Agent: Agentic Reinforcement Learning for Automatic Heuristic Design
AHD Agent trains a 4B-parameter LLM via agentic RL to actively use tools for automatic heuristic design, matching or exceeding larger baselines across eight domains with fewer evaluations.
-
Value-Decomposed Reinforcement Learning Framework for Taxiway Routing with Hierarchical Conflict-Aware Observations
CaTR applies value-decomposed RL with hierarchical conflict-aware observations to achieve better safety-efficiency trade-offs than planning, optimization, and standard RL baselines in a realistic airport taxiway simulation.
-
The Attacker in the Mirror: Breaking Self-Consistency in Safety via Anchored Bipolicy Self-Play
Anchored Bipolicy Self-Play trains role-specific LoRA adapters on a frozen base model to break self-consistency collapse in self-play red-teaming, yielding up to 100x parameter efficiency and stronger safety on Qwen2.5 models.
-
Agentick: A Unified Benchmark for General Sequential Decision-Making Agents
Agentick is a new benchmark for sequential decision-making agents that evaluates RL, LLM, VLM, hybrid, and human approaches across 37 tasks and finds no single method dominates.
-
Randomness is sometimes necessary for coordination
Structured per-agent randomness via ranked masking in attention allows symmetric agents to break ties and coordinate, achieving perfect success on symmetric tasks where deterministic policies fail and enabling zero-shot transfer across team sizes.
-
Post Reasoning: Improving the Performance of Non-Thinking Models at No Cost
Post-Reasoning boosts LLM accuracy by reversing the usual answer-after-reasoning order, delivering mean relative gains of 17.37% across 117 model-benchmark pairs with zero extra cost.
-
Aligning with Your Own Voice: Self-Corrected Preference Learning for Hallucination Mitigation in LVLMs
AVES-DPO mitigates hallucinations in LVLMs by creating in-distribution preference pairs through the model's self-correction, outperforming baselines with only 5.2k samples.
-
Agentic Adversarial Rewriting Exposes Architectural Vulnerabilities in Black-Box NLP Pipelines
A two-agent adversarial rewriting framework achieves 20-40% evasion rates against LLM-based misinformation detectors under strict black-box constraints with binary feedback only, far outperforming prior methods and linking success to specific architectural properties.
-
Waking Up Blind: Cold-Start Optimization of Supervision-Free Agentic Trajectories for Grounded Visual Perception
SPECTRA enables supervision-free bootstrapping of agentic capabilities in SVLMs via cascaded tool rollout alignment, multi-objective rewards, and the TIU metric, yielding up to 5% higher task accuracy and 9% better tool efficiency.
-
IG-Search: Step-Level Information Gain Rewards for Search-Augmented Reasoning
IG-Search computes step-level information gain rewards from policy probabilities to improve credit assignment in RL training for search-augmented QA, yielding 1.6-point gains over trajectory-level baselines on multi-hop tasks.
-
Advantage-Guided Diffusion for Model-Based Reinforcement Learning
Advantage-guided diffusion (SAG and EAG) steers sampling in diffusion world models to higher-advantage trajectories, enabling policy improvement and better sample efficiency on MuJoCo tasks.
-
PRIME: Training Free Proactive Reasoning via Iterative Memory Evolution for User-Centric Agent
PRIME enables agents to proactively reason in user-centric tasks by iteratively evolving structured memories from interaction trajectories without gradient-based training.
-
Neural Assistive Impulses: Synthesizing Exaggerated Motions for Physics-based Characters
A hybrid neural policy operating in impulse space enables physics-based characters to track exaggerated, dynamically infeasible motions that standard DRL methods cannot stabilize.
-
RL-VLA$^3$: A Flexible and Asynchronous Reinforcement Learning Framework for VLA Training
RL-VLA³ is an asynchronous RL framework for VLA training that delivers up to 85.2% higher throughput than synchronous baselines while preserving identical sample efficiency and scaling to 256 GPUs.
-
Small Generalizable Prompt Predictive Models Can Steer Efficient RL Post-Training of Large Reasoning Models
GPS trains a small model on optimization history to predict prompt difficulty and select intermediate-difficulty diverse batches, yielding better training efficiency, final performance, and test-time allocation than baselines on reasoning benchmarks.
-
CORE: Concept-Oriented Reinforcement for Bridging the Definition-Application Gap in Mathematical Reasoning
CORE is a concept-oriented RL method that synthesizes quizzes, injects concept snippets into rollouts, and reinforces conceptual trajectories to close the gap between restating definitions and applying them in math problems.
-
MixGRPO: Unlocking Flow-based GRPO Efficiency with Mixed ODE-SDE
MixGRPO speeds up GRPO for flow-based image generators by restricting SDE sampling and optimization to a sliding window while using ODE elsewhere, cutting training time by up to 71% with better alignment performance.
-
One Step is Enough: Multi-Agent Reinforcement Learning based on One-Step Policy Optimization for Order Dispatch on Ride-Sharing Platforms
OSPO trains optimal order dispatch policies for homogeneous AV fleets using only one-step group rewards, outperforming GRPO on a real ride-hailing dataset.
-
Universal Reasoner: A Single, Composable Plug-and-Play Reasoner for Frozen LLMs
UniR is a composable reasoning module trained with verifiable rewards and added to frozen LLMs via logit summation, enabling modular composition and weak-to-strong generalization across tasks and model sizes.
-
R1-VL: Learning to Reason with Multimodal Large Language Models via Step-wise Group Relative Policy Optimization
R1-VL uses StepGRPO with rule-based StepRAR and StepRVR rewards to let MLLMs learn step-by-step reasoning beyond imitation of positive paths.
-
Sycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Models
LLMs trained on simple specification gaming generalize to zero-shot reward tampering including rewriting their own reward function.
-
Learning Interactive Real-World Simulators
UniSim learns a universal real-world simulator from orchestrated diverse datasets, enabling zero-shot deployment of policies trained purely in simulation.
-
Voyager: An Open-Ended Embodied Agent with Large Language Models
Voyager achieves superior lifelong learning in Minecraft by combining an automatic exploration curriculum, a library of executable skills, and iterative LLM prompting with environment feedback, yielding 3.3x more unique items and 15.3x faster milestone unlocks than prior methods while generalizing技能
-
Mastering Diverse Domains through World Models
DreamerV3 uses world models and robustness techniques to solve over 150 tasks across domains with a single configuration, including Minecraft diamond collection from scratch.
-
SDR: Set-Distance Rewards for Radiology Report Generation
Set-to-set distances on sentence embeddings provide a permutation-invariant reward signal that improves GRPO training and enables efficient test-time scaling for vision-language models generating chest X-ray reports.
-
Robust Shielding for Safe Reinforcement Learning
A sound and optimal shielding method for robust MDPs ensures LTL safety under worst-case transitions and combines with PAC sampling to produce minimally restrictive shields for learned models.
-
HADT: A Heterogeneous Multi-Agent Differential Transformer for Autonomous Earth Observation Satellite Cluster
Proposes HADT, a heterogeneous multi-agent differential transformer with relational observations-actions tokenization for model-free RL-based autonomous resource management in EO satellite clusters, claiming gains over baselines and adaptability to cluster size changes.
-
Quantifying and Optimizing Simplicity via Polynomial Representations
Polynomial representations yield an effective-degree simplicity metric that predicts generalization across tasks and serves as a differentiable regularizer improving performance in classification and RL.
-
ReasonLight: A Multimodal Foundation Model-Enhanced Reinforcement Learning Framework for Zero-Shot Traffic Signal Control
ReasonLight uses multimodal foundation models to refine RL-proposed traffic signal phases based on camera images and sensor data, enabling zero-shot adaptation to unseen events such as emergency vehicle priority.
-
One Policy, Infinite NPCs: Persona-Traceable Shared RL Policies for Scalable Game Agents
pcsp is a shared RL policy using LLM persona embeddings, low-rank projection, and PPO+InfoNCE+KL training that delivers 17x above-chance zero-shot persona identification and 22x faster inference on a 300-persona benchmark.
-
CLORE: Content-Level Optimization for Reasoning Efficiency
CLORE augments correct on-policy rollouts by deleting repetitive and irrelevant segments then optimizes with auxiliary DPO to improve accuracy-efficiency trade-off on math benchmarks.
-
Attention-Guided Reward for Reinforcement Learning-based Jailbreak against Large Reasoning Models
An attention-guided RL reward combined with diverse persuasion strategies produces higher attack success rates against large reasoning models than prior jailbreak methods.
-
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.
-
SD-Search: On-Policy Hindsight Self-Distillation for Search-Augmented Reasoning
SD-Search derives step-level supervision for search queries in reasoning agents via on-policy hindsight self-distillation using the policy as both student and teacher.
-
PAIR: Prefix-Aware Internal Reward Model for Multi-Turn Agent Optimization
PAIR combines a hidden-state probe with an attention correction to deliver robust step-level rewards for GRPO-based optimization of multi-turn LLM agents, achieving high AUROC on contaminated trajectories at low cost.