AI reviews for all 22,977 AAAI-26 papers were preferred by authors and PC members over human reviews on accuracy and suggestions and outperformed baselines at spotting weaknesses.
mega hub Mixed citations
DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
Mixed citation behavior. Most common role is background (53%).
1640
Pith papers citing it
Background
53% of classified citations
abstract
Mathematical reasoning poses a significant challenge for language models due to its complex and structured nature. In this paper, we introduce DeepSeekMath 7B, which continues pre-training DeepSeek-Coder-Base-v1.5 7B with 120B math-related tokens sourced from Common Crawl, together with natural language and code data. DeepSeekMath 7B has achieved an impressive score of 51.7% on the competition-level MATH benchmark without relying on external toolkits and voting techniques, approaching the performance level of Gemini-Ultra and GPT-4. Self-consistency over 64 samples from DeepSeekMath 7B achieves 60.9% on MATH. The mathematical reasoning capability of DeepSeekMath is attributed to two key factors: First, we harness the significant potential of publicly available web data through a meticulously engineered data selection pipeline. Second, we introduce Group Relative Policy Optimization (GRPO), a variant of Proximal Policy Optimization (PPO), that enhances mathematical reasoning abilities while concurrently optimizing the memory usage of PPO.
hub tools
citation-role summary
background 188
method 124
baseline 14
other 5
dataset 4
citation-polarity summary
claims ledger
- abstract Mathematical reasoning poses a significant challenge for language models due to its complex and structured nature. In this paper, we introduce DeepSeekMath 7B, which continues pre-training DeepSeek-Coder-Base-v1.5 7B with 120B math-related tokens sourced from Common Crawl, together with natural language and code data. DeepSeekMath 7B has achieved an impressive score of 51.7% on the competition-level MATH benchmark without relying on external toolkits and voting techniques, approaching the performance level of Gemini-Ultra and GPT-4. Self-consistency over 64 samples from DeepSeekMath 7B achieve
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
Noisy expert imitation learning requires exponential samples for offline methods but polynomial for a variant of on-policy distillation under a noise condition.
Introduces textual belief states and factorized GRPO to enforce strict latent state mediation in text-based world models, yielding preserved prediction accuracy with large gains in representation quality and rollout performance on TextWorld and ScienceWorld.
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.
EGLR adds a deterministic layer-recursion axis gated by entropy that is complementary to temperature sampling, raising joint oracle accuracy on MATH-500 from 83.4% to 91.6% for a 3B model.
Introduces the TVR active viewpoint-matching task and TVRBench indoor simulation benchmark, where foundation models start at low single-digit success rates but reach 51.4% after visual-action SFT and multi-turn GRPO post-training.
L-layer transformers under Log-ICoT curriculum provably learn k-parity with poly(n) samples and log k stages, matching explicit CoT efficiency without inference overhead.
Continual Harness automates online self-improvement for foundation-model embodied agents by refining prompts, sub-agents, skills, and memory within one run, cutting button-press costs on Pokemon Red and Emerald and closing much of the gap to expert harnesses.
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.
STARE uses step-wise RL to attack multimodal models, achieving 68% higher attack success rate while revealing that adversarial optimization concentrates conceptual toxicity early and detail toxicity late in the generation trajectory.
MLLMs exhibit a Mirage effect by bypassing circuit diagrams in favor of header semantics for Verilog generation; VeriGround with identifier anonymization and D-ORPO training reaches 46% Functional Pass@1 while refusing blank images at >92%.
S1-VL combines structured scientific reasoning with iterative image manipulation via code execution to reach state-of-the-art results on visual and scientific reasoning benchmarks.
EVE enables verifiable self-evolution of MLLMs by using a Challenger-Solver architecture to generate dynamic executable visual transformations that produce VQA problems with absolute execution-verified ground truth.
MedPRMBench is the first fine-grained benchmark for process reward models in medical reasoning, featuring 6500 questions, 13000 chains, 113910 step labels, and a baseline that improves downstream QA accuracy by 3.2-6.7 points.
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.
GIANTS-4B, trained with RL on a new 17k-example benchmark of parent-to-child paper insights, achieves 34% relative improvement over gemini-3-pro in LM-judge similarity and is rated higher-impact by a citation predictor.
SARL rewards reasoning topology to improve label-free RL, outperforming baselines with gains up to 44.7% on math and 34.6% on open-ended tasks while maintaining more stable training.
SEVerA uses Formally Guarded Generative Models and a three-stage Search-Verification-Learning process to synthesize self-evolving agents that satisfy hard formal constraints while improving task performance.
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.
RLCracker is a reinforcement learning attack that erases LLM watermarks at 98.5% success rate with minimal data and generalizes across ten schemes and multiple model sizes.
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.
DeepMath-103K is a new 103K-problem mathematical dataset with high difficulty, rigorous decontamination, and verifiable answers to support RL training of language-model reasoning.
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.
QVal is a new evaluation framework that directly measures dense supervision quality via Q-alignment to a reference policy, showing simple prompting baselines outperform 21 other methods across environments and models.
citing papers explorer
-
Flow-GRPO: Training Flow Matching Models via Online RL
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.
-
GeoVista: Visually Grounded Active Perception for Ultra-High-Resolution Remote Sensing Understanding
GeoVista introduces a planning-driven active perception framework with global exploration plans, branch-wise local inspection, and explicit evidence tracking to achieve state-of-the-art results on ultra-high-resolution remote sensing benchmarks.
-
ReTool-Video: Recursive Tool-Using Video Agents with Meta-Augmented Tool Grounding
ReTool-Video uses a 134-tool meta-augmented library and recursive grounding to translate abstract video intents into fine-grained multimodal operations, outperforming baselines on MVBench, MLVU, and Video-MME.
-
Boosting Omni-Modal Language Models: Staged Post-Training with Visually Debiased Evaluation
Visual debiasing of omni-modal benchmarks combined with staged post-training lets a 3B model match or exceed a 30B model without a stronger teacher.
-
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.
-
Rebellious Student: Reversing Teacher Signals for Reasoning Exploration with Self-Distilled RLVR
RLRT augments GRPO by reinforcing tokens on correct student rollouts that the teacher would not have predicted, outperforming standard self-distillation and exploration baselines on Qwen3 models.
-
The Extrapolation Cliff in On-Policy Distillation of Near-Deterministic Structured Outputs
On-policy distillation has an extrapolation cliff at closed-form lambda*(p,b,c) set by teacher modal probability, warm-start mass, and clip strength, past which training shifts from format-preserving to format-collapsing.
-
DUET: Optimize Token-Budget Allocation for Reinforcement Learning with Verifiable Rewards
DUET improves RLVR by allocating tokens across both prompt selection and rollout length, outperforming full-budget baselines even when using only half the tokens.
-
GazeVLM: Active Vision via Internal Attention Control for Multimodal Reasoning
GazeVLM introduces internal gaze tokens that allow VLMs to dynamically suppress irrelevant visual features and simulate foveal attention for improved high-resolution multimodal reasoning.
-
NoiseGate: Learning Per-Latent Timestep Schedules as Information Gating in World Action Models
NoiseGate learns per-latent timestep schedules as an information-gating policy in diffusion-based world action models, yielding consistent gains on RoboTwin manipulation tasks.
-
MemCompiler: Compile, Don't Inject -- State-Conditioned Memory for Embodied Agents
MemCompiler reframes memory use as state-conditioned compilation, delivering relevant guidance via text and latent channels to improve embodied agent performance up to 129% and cut latency 60% versus static injection.
-
Teaching Language Models to Think in Code
ThinC trains small models to reason primarily in code rather than natural language, outperforming tool-integrated baselines and even larger models on competition math benchmarks.
-
Weblica: Scalable and Reproducible Training Environments for Visual Web Agents
Weblica scales RL training for visual web agents by building thousands of reproducible environments through HTTP caching for stable replays and LLM synthesis from real sites, yielding an 8B model that beats similar open baselines on navigation benchmarks.
-
Enhanced LLM Reasoning by Optimizing Reward Functions with Search-Driven Reinforcement Learning
Iterative search over reward functions with ranked feedback in GRPO training improves LLM math reasoning, achieving F1 of 0.795 on GSM8K versus 0.609 for baseline.
-
HP-Edit: A Human-Preference Post-Training Framework for Image Editing
HP-Edit introduces a post-training framework and RealPref-50K dataset that uses a VLM-based HP-Scorer to align diffusion image editing models with human preferences, improving outputs on Qwen-Image-Edit-2509.
-
Guiding Distribution Matching Distillation with Gradient-Based Reinforcement Learning
GDMD replaces raw-sample rewards with distillation-gradient rewards in RL-guided diffusion distillation, yielding 4-step models that surpass their multi-step teachers on GenEval and human preference metrics.
-
SAT: Sequential Agent Tuning for Coordinator Free Plug and Play Multi-LLM Training with Monotonic Improvement Guarantees
SAT trains multi-LLM teams with sequential block updates to deliver monotonic gains and plug-and-play model swaps that provably improve performance bounds.
-
Visual Preference Optimization with Rubric Rewards
rDPO uses offline-built rubrics to generate on-policy preference data for DPO, raising benchmark scores in visual tasks over outcome-based filtering and style baselines.
-
RLSpoofer: A Lightweight Evaluator for LLM Watermark Spoofing Resilience
RLSpoofer trains a 4B model on 100 watermarked paraphrase pairs to spoof PF watermarks at 62% success rate, far exceeding baselines trained on up to 10,000 samples.
-
MMR-AD: A Large-Scale Multimodal Dataset for Benchmarking General Anomaly Detection with Multimodal Large Language Models
MMR-AD is a new benchmark dataset showing that current generalist MLLMs lag industrial needs for anomaly detection, with Anomaly-R1 delivering better results through reasoning and RL.
-
Learning Vision-Language-Action World Models for Autonomous Driving
VLA-World improves autonomous driving by using action-guided future image generation followed by reflective reasoning over the imagined scene to refine trajectories.
-
RemoteAgent: Bridging Vague Human Intents and Earth Observation with RL-based Agentic MLLMs
RemoteAgent uses RL fine-tuning on VagueEO to align MLLMs for vague EO intent recognition, handling simple tasks internally and routing dense predictions to tools via Model Context Protocol.
-
MinerU2.5-Pro: Pushing the Limits of Data-Centric Document Parsing at Scale
A fixed 1.2B model trained via diversity-aware sampling, cross-model verification, annotation refinement, and progressive stages achieves new state-of-the-art document parsing accuracy of 95.69 on OmniDocBench v1.6.
-
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.
-
Video-R1: Reinforcing Video Reasoning in MLLMs
Video-R1 uses temporal-aware RL and mixed datasets to boost video reasoning in MLLMs, with a 7B model reaching 37.1% on VSI-Bench and surpassing GPT-4o.
-
ToolCUA: Towards Optimal GUI-Tool Path Orchestration for Computer Use Agents
ToolCUA introduces a trajectory scaling pipeline and staged RL to optimize GUI-tool switching, reaching 46.85% accuracy on OSWorld-MCP for a 66% relative gain over baseline.
-
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.
-
Reinforcing VLAs in Task-Agnostic World Models
RAW-Dream disentangles world-model learning from task data by using a pre-trained task-agnostic world model and VLM rewards, with dual-noise filtering, to enable zero-shot VLA adaptation in simulation and real settings.
-
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.
-
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.
-
GEAR: Granularity-Adaptive Advantage Reweighting for LLM Agents via Self-Distillation
GEAR adaptively reweights GRPO advantages in LLM RL by using divergence spikes from self-distillation to define semantic segments and modulate local credit.
-
Seir\^enes: Adversarial Self-Play with Evolving Distractions for LLM Reasoning
Seirênes trains LLMs via adversarial self-play to generate and overcome evolving distractions, producing gains of 7-10 points on math reasoning benchmarks and exposing blind spots in larger models.
-
Anti-Self-Distillation for Reasoning RL via Pointwise Mutual Information
Anti-Self-Distillation reverses self-distillation signals via PMI to fix overconfidence on structural tokens, matching GRPO baseline accuracy 2-10x faster with up to 11.5 point gains across 4B-30B models.
-
Drop the Act: Probe-Filtered RL for Faithful Chain-of-Thought Reasoning
ProFIL trains an activation probe on a frozen base model to zero advantages on theatrical post-commitment rollouts in GRPO, cutting theater 11-100%, raising faithful fractions, and shortening chains 4-19% without accuracy loss.
-
Internalizing Curriculum Judgment for LLM Reinforcement Fine-Tuning
METIS internalizes curriculum judgment in LLM reinforcement fine-tuning by predicting within-prompt reward variance via in-context learning and jointly optimizing with a self-judgment reward, yielding superior performance and up to 67% faster convergence across math, code, and agent benchmarks.
-
Shepherd: Enabling Programmable Meta-Agents via Reversible Agentic Execution Traces
Shepherd provides a reversible execution trace substrate for LLM agents that enables meta-agents to inspect and transform runs, yielding reported gains on coding and terminal benchmarks via supervision, counterfactual repair, and RL credit assignment.
-
ProteinOPD: Towards Effective and Efficient Preference Alignment for Protein Design
ProteinOPD uses token-level on-policy distillation from multiple preference-specific teacher models into a shared student to balance competing objectives in protein design, delivering gains on targets without losing designability and an 8x speedup over RL baselines.
-
Efficient LLM Reasoning via Variational Posterior Guidance with Efficiency Awareness
VPG-EA applies variational posterior guidance and efficiency-aware distillation to compress LLM reasoning chains while preserving performance.
-
DeltaRubric: Generative Multimodal Reward Modeling via Joint Planning and Verification
DeltaRubric decomposes multimodal preference evaluation into self-generated planning and verification steps within a single model, producing large accuracy improvements on VL-RewardBench via multi-role reinforcement learning.
-
Reinforcing Multimodal Reasoning Against Visual Degradation
ROMA improves MLLM robustness to seen and unseen visual corruptions by +2.3-2.4% over GRPO on seven reasoning benchmarks while matching clean accuracy.
-
Iterative Critique-and-Routing Controller for Multi-Agent Systems with Heterogeneous LLMs
A critique-and-routing controller cast as a finite-horizon MDP with policy-gradient optimization outperforms one-shot routing baselines on reasoning benchmarks while using the strongest agent for under 25% of calls.
-
Pushing Biomolecular Utility-Diversity Frontiers with Supergroup Relative Policy Optimization
SGRPO is a GRPO-style framework that constructs set-level diversity rewards via supergroup sampling and leave-one-out redistribution to expand the utility-diversity Pareto frontier in biomolecular design tasks.
-
AIPO: Learning to Reason from Active Interaction
AIPO adds active multi-agent consultation (Verify, Knowledge, Reasoning agents) plus custom importance sampling to RLVR training so LLMs expand their reasoning boundary and then operate without the agents.
-
Proxy3D: Efficient 3D Representations for Vision-Language Models via Semantic Clustering and Alignment
Proxy3D generates efficient 3D proxy representations via semantic clustering from video frames and aligns them to VLMs through multi-stage training on the new SpaceSpan dataset, achieving competitive performance on 3D VQA, grounding, and spatial benchmarks with shorter sequences.
-
Beyond Pairs: Your Language Model is Secretly Optimizing a Preference Graph
GraphDPO generalizes pairwise DPO to a graph-structured Plackett-Luce objective over DAGs induced by rollout rankings, enforcing transitivity with linear complexity and recovering DPO as a special case.
-
Video Understanding Reward Modeling: A Robust Benchmark and Performant Reward Models
Introduces VURB benchmark and VUP-35K dataset to train discriminative and generative video reward models that achieve SOTA performance on VURB and VideoRewardBench.
-
From Synthetic to Real: Toward Identity-Consistent Makeup Transfer with Synthetic and Real Data
The work creates identity-consistent synthetic makeup data via ConsistentBeauty and adapts models to real images using reinforcement learning in RealBeauty, achieving better identity preservation and real-world performance than prior methods.
-
SOD: Step-wise On-policy Distillation for Small Language Model Agents
SOD reweights on-policy distillation strength step-by-step using divergence to stabilize tool use in small language model agents, yielding up to 20.86% gains and 26.13% on AIME 2025 for a 0.6B model.
-
ReasonEdit: Towards Interpretable Image Editing Evaluation via Reinforcement Learning
ReasonEdit uses a new CoT dataset and reinforcement learning to produce interpretable, human-aligned evaluations of text-guided image edits.
-
RRCM: Ranking-Driven Retrieval over Collaborative and Meta Memories for LLM Recommendation
RRCM trains an LLM to dynamically retrieve from collaborative and meta memories using group relative policy optimization driven by final top-k recommendation quality.