Hugging Face Daily Papers
500 articles archived · Visit source ↗ · RSS
-
Hugging Face Daily Papers research 3d ago
MARS: Multi-Specialist LLM Relay System for Competitive Programming
Abstract MARS uses retrieval-augmented specialist agents for algorithmic topics to iteratively generate, test, and refine C++ solutions, improving competitive programming pass rates with lower cost. Generated by thinkingmachines/Inkling-Small Large Language Models excel at code…
31 -
When "Must" Becomes "Maybe": Constraint Weakening in LLM Agent Workflows
Abstract Multi-stage LLM workflows lose operational constraints when intermediate artifacts transform binding prerequisites into non-binding context, causing safety failures despite preserved content. Generated by thinkingmachines/Inkling-Small Large language model (LLM) agents…
32 -
Autonomous Mathematical Discovery in an Open-World Multi-Agent Environment
Abstract We study autonomous mathematical discovery in the Station, an open-world multi-agent environment in which AI agents from different model families pursue a shared research goal without a central coordinator or scripted pipeline. Agents choose their own research…
16 -
SecOPD: Mitigating Adaptive Prompt Injections by On-Policy Distillation
Abstract SecOPD improves defense against adaptive prompt injection by using token-level feedback during fine-tuning, sharply reducing attack success rates on language models. Generated by thinkingmachines/Inkling-Small Prompt injection is listed as the \#1 threat to AI agents.…
18 -
AgentRoom: Concurrent Multi-Agent Coding in a CRDT-Backed Shared Workspace
Abstract AgentRoom enables concurrent multi-agent coding through real-time collaborative editing and shared filesystem coordination, reducing task abandonment and improving consistency compared to solo or uncoordinated parallel approaches. Generated by…
5 -
Automata from Agent Traces: Failure and Next-Step Prediction
Abstract LLM agent traces are compressed into compact finite-state machines that enable accurate next-step and failure prediction for safety auditing and runtime monitoring. Generated by thinkingmachines/Inkling-Small LLM-based agents execute multi-step tasks, but their…
23 -
MoTE: Mixture of Task Experts for Multi-Task Video Understanding
Abstract MoTE replaces dense decoder feed-forward networks with task-specific experts routed by sample-level task labels, improving multi-task video-language accuracy with sparse, interpretable computation. Generated by thinkingmachines/Inkling-Small Procedural video-language…
13 -
DREAM Technical Report
Abstract DREAM introduces an agentic meta-control layer over industrial recommender pipelines that uses intent reasoning and dual-loop optimization to improve session-level recommendations without replacing existing models. Generated by thinkingmachines/Inkling-Small Industrial…
25 -
Latent Action as Intention Enables Efficient Future Imagination for World Action Models
Abstract LAWA improves robot control by using compact latent actions to retain efficient future imagination without generating observations, achieving strong performance with lower latency. Generated by thinkingmachines/Inkling-Small World action models (WAMs) improve robot…
6 -
CyberFactory: Scaling Cyber Security Capabilities with Instances from the Wild
Abstract CyberFactory is an open-source framework that builds agentic training data from real vulnerabilities to train Aegis, improving open-weight cybersecurity performance across proof-of-concept generation, patching, and question answering. Generated by…
13 -
CAFE: Self-Improving Search Agents Need Co-Evolving Feedback
Abstract CAFE couples a search agent and critic via shared parameters to learn in-trajectory corrective feedback, improving search performance and reducing hallucinations across benchmarks. Generated by thinkingmachines/Inkling-Small Outcome-supervised search agents learn when…
19 -
Meta^n: Recursive Self-Improvement through Emergent Depth
Abstract Meta^n recursively applies a fixed meta-operation to growing inputs, building deeper reasoning layers that improve self-improving LLM agents without destabilizing the system. Generated by thinkingmachines/Inkling-Small Self-improving LLM agents refine answers, not the…
16 -
On-policy Distillation with Verifiable Reward
Abstract OPDVR integrates on-policy distillation with verifiable rewards via a ReLU-gated implicit reward reformulation, improving reasoning performance without extra hyperparameters. Generated by thinkingmachines/Inkling-Small Reinforcement Learning with Verifiable Rewards…
35 -
LAION-BVD: A 10-Million-Hour Open Video Dataset for Multimodal Pre-training
Abstract LAION-BVD is a large-scale open video dataset enabling multimodal pre-training across video, audio, and image modalities with synthetic captions and strong benchmark performance. Generated by thinkingmachines/Inkling-Small We present LAION-BVD, a large-scale open video…
22 -
Game2World Engine: Unlocking In-the-Wild Gameplay Videos for World Model Training
Abstract A framework for removing gameplay UI from videos enables cleaner training data for video world models, improving reward metrics and outperforms mask-based removal methods. Generated by thinkingmachines/Inkling-Small Video games provide a scalable source of training data…
11 -
Annotations as Rollouts: Efficient and Scalable Reinforcement Learning for Video MLLMs
Abstract OraRL improves reinforcement learning post-training for video multimodal language models by integrating oracle rollouts with decoupled advantage estimation and sign-balanced pruning, achieving higher sample efficiency and scalability without chain-of-thought generation.…
17 -
From Seeing to Acting: Smart Glasses as First-Person Intelligence Platforms
Abstract Smart glasses are surveyed as unified first-person intelligence platforms requiring sustained perception-state-interaction-action loops across constrained hardware and diverse applications. Generated by thinkingmachines/Inkling-Small Smart glasses are evolving from…
14 -
WarpSAC: Towards the Pinnacle of Scalable Off-policy RL by Rethinking Exploration and Exploitation
Abstract Off-policy reinforcement learning stabilizers vary with data availability, motivating regime-aware algorithms that adapt normalization and Q-function clipping to improve efficiency across CPU and GPU-parallel training. Generated by thinkingmachines/Inkling-Small…
19 -
WeMM-Embedding: WeChat Multi-Modal Embedding Technical Report
Abstract WeMM-Embedding is a family of universal multimodal embedding models that align text, images, videos, and interleaved inputs in a shared space, achieving state-of-the-art retrieval and recommendation performance across public benchmarks and large-scale WeChat…
15 -
TorchMorph: CUDA-accelerated Morphological Transforms
Abstract TorchMorph is a PyTorch extension providing GPU-accelerated morphological and distance-transform operators across up to eight dimensions with a SciPy-compatible API. Generated by thinkingmachines/Inkling-Small Morphological transforms are long-standing tools for shape…
14 -
Length-Adaptive Decoding for Masked Diffusion Machine Translation
Abstract Entropy-Valley selects target lengths for masked diffusion translation by scoring predictive entropy, improving adequacy and showing length choice matters more than unmasking order. Generated by thinkingmachines/Inkling-Small Machine translation tests masked diffusion…
36 -
Best Practice Critic Optimization
Abstract BPCO stabilizes critic-based reinforcement learning for language models by combining bounded value predictions, Monte Carlo targets, and adaptive advantage estimation, matching group-based methods with single-response sampling. Generated by…
6 -
AutoSaddler: Automatic Harness Optimization with Durable Updates from Agent Execution Traces
Abstract AutoSaddler automatically improves LLM agent harnesses via offline failure-driven optimization, boosting performance on long-horizon benchmarks. Generated by thinkingmachines/Inkling-Small LLM agents remain unreliable on long-horizon tasks, where small local failures…
26 -
On-Policy Self-Distillation in Diffusion Models
Abstract DiffusionOPSD uses on-policy self-distillation to turn image-level rewards into explicit intermediate targets for diffusion models, improving alignment efficiency and enabling separate analysis of target construction and policy fitting. Generated by…
16 -
Recursive Experiential-Working Memory Evolution for Long-Horizon Agent Harnesses
Abstract Recuris introduces a recursive memory architecture that tracks progress and guides skill selection to improve long-horizon agent success through localized, validation-gated updates. Generated by thinkingmachines/Inkling-Small Recursive self-improvement (RSI) remains…
25 -
GigaBrain-0.7: Scaling Embodied Foundation Models to Emergent Capabilities with a Three-System Architecture
Abstract GigaBrain-0.7 is a vision-language-action model that improves embodied generalization via a three-system architecture, large-scale heterogeneous pretraining, and joint alignment training. Generated by thinkingmachines/Inkling-Small Vision-language-action (VLA) models…
31 -
Gated Recurrent Transformers: Expressive Depth through Recurrent Modulation in Transformers
Abstract A gated recurrent transformer reuses a shared core across depth with adaptive update gates, achieving comparable or better quality than deeper models with far fewer parameters and lower memory. Generated by thinkingmachines/Inkling-Small Scaling transformer language…
12 -
GameXpert-Bench: How Far Are Coding Agents from Expert Game Development?
Abstract GameXpert-Bench evaluates coding agents across three game development stages—generation, repair, and optimization—using interactive and behavioral tests to reveal strengths in building playable foundations and weaknesses in defect discovery and regression preservation.…
20 -
From Generation to Simulation: How Far Are World Models from Being True Simulators?
Abstract Generative world models are evaluated against traditional simulators across eight capabilities, revealing gaps in physical guarantees, state feedback, and long-horizon stability despite progress in interaction and controllability. Generated by…
11 -
RIBOSPAN: A Long-Context RNA Foundation Model for Versatile RNA Modeling
Abstract RIBOSPAN is a large bidirectional RNA foundation model pretrained on up to 10,240 nucleotides that enables high-resolution full-transcript modeling, strong long-context representations, and discrete-diffusion-based mRNA generation and redesign. Generated by…
29 -
Beyond the Stability-Exploration Dilemma: Environmental Regularization for LLM Policy Optimization
Abstract ERPO replaces action-side policy regularization with input-side query distribution control to stabilize reinforcement learning for language models while preserving response exploration. Generated by thinkingmachines/Inkling-Small Policy optimization (PO) for Large…
4 -
EchoWM: Open and Enterable Omnimodal World Models
Abstract EchoWM is an omnimodal world model that generates synchronized high-resolution video, sound, music, and speech while following continuous 6-DoF navigation trajectories across first- and third-person views. Generated by thinkingmachines/Inkling-Small We present EchoWM,…
17 -
One Success Isn't Reliability: Thinkingbox, a Sandbox and Benchmark for Agents in Stateful Business Workflows
Abstract Recent agent benchmarks increasingly ground evaluation in executable environments, from code repair to web navigation, app APIs, and function calling. Yet completing consequential work beyond code requires more than producing a plausible response or valid tool call:…
33 -
One Polluted Page Is Enough: Evaluating Web Content Pollution in LLM Recommenders
Abstract Search-augmented LLM recommenders are highly vulnerable to web content polluted by generative engine optimization, frequently promoting fake products despite reasoning and defenses. Generated by thinkingmachines/Inkling-Small Search-augmented LLMs increasingly mediate…
26 -
Block3D: Efficient Text-to-3D Generation via Block-Wise Diffusion
Abstract Block3D accelerates text-to-3D generation by using block-wise diffusion with confidence-guided correction to reduce inference time while preserving geometric fidelity. Generated by thinkingmachines/Inkling-Small While text-to-3D generation has advanced rapidly,…
13 -
MobilePA-Bench: Benchmarking Mobile Planner Agents on Complex Real-World Tasks
Abstract MobilePA-Bench is an interactive sandbox benchmark that evaluates mobile planning agents on tool-calling, sub-agent collaboration, memory usage, and composite skill invocation under real runtime constraints. Generated by thinkingmachines/Inkling-Small As on-device LLM…
13 -
Better Retrieval, Worse Robustness:How Multi-hop RAG Amplifies Upstream ASR Errors
Abstract Retrieval-augmented generation extensions amplify automatic speech recognition errors in spoken multi-hop question answering, primarily through corrupted query entities. Generated by thinkingmachines/Inkling-Small Speech-based applications pass spoken queries through…
12 -
Task-CoEvolve: Efficient Harness Optimization via Adaptive Validation Task Selection
Abstract Task-CoEvolve improves LLM harness optimization by adaptively selecting validation tasks and estimating full-set performance from partial evaluations, cutting evaluation costs by 80%. Generated by thinkingmachines/Inkling-Small We present a novel approach to efficient…
30 -
ARC: Fair Relative Advantage Comparison in Open-Ended Real-World Interaction
Abstract ARC improves fairness in group-based reinforcement learning for open-ended agents by conditioning rollout comparisons on strategy, enabling more context-appropriate behavior in responsive user-agent interaction. Generated by thinkingmachines/Inkling-Small Open-ended…
15 -
Unlocking the Potential of Image Editing via Concept Scaling and Dense Supervision
Abstract A hierarchical taxonomy and dense supervision strategy improve diffusion-based image editing through fine-grained concepts, large-scale paired data, and granular evaluation. Generated by thinkingmachines/Inkling-Small Existing image editing frameworks predominantly…
17 -
ReWorld: An Interactive World Model with Long-Horizon Memory
Abstract ReWorld separates short-horizon control and long-horizon memory during training, then bounds both at inference via mixed attention windows, a pose-indexed landmark bank, and distribution-matching LoRA distillation to enable real-time interactive world modeling with…
8 -
Prime Agent: A Self-Improving RLM Harness
Abstract Prime Agent is an open-source harness that uses recursive subagents, persistent computation, and agent-to-agent coordination to extend language models' long-horizon capabilities across coding and reasoning tasks. Generated by thinkingmachines/Inkling-Small Language…
22 -
Beyond Imitation: Filtering On-Policy Distillation by Reasoning Progress
Abstract R2-OPD improves on-policy distillation by filtering teacher rewards that conflict with reasoning progress via within-trajectory ranking comparisons. Generated by thinkingmachines/Inkling-Small On-policy distillation (OPD) has emerged as an effective framework for…
25 -
TileMix: Tile-Centric Mixed-Precision Attention for LLM Inference Acceleration
Abstract TileMix routes attention score tiles to mixed FP16 or INT8 precision within fused dense attention, recovering long-context accuracy while improving prefill throughput without retraining. Generated by thinkingmachines/Inkling-Small Long-context prefill in large language…
5 -
Same Agent, Different Answers: A Repeat-Aware Audit of Corpus-Induced Answer Churn in Retrieval-Augmented QA
Abstract Retrieval-augmented QA systems can exhibit hidden answer churn during index updates without noticeable accuracy changes, motivating compatibility audits alongside utility evaluations. Generated by thinkingmachines/Inkling-Small A retrieval-augmented QA system can return…
21 -
Apodex 1.1: Scaling Agentic Intelligence for Complex Work
Abstract Apodex 1.1 improves sustained, verifiable progress on complex real-world tasks by scaling executable environments and training agents to coordinate long-horizon work with state maintenance and recovery. Generated by thinkingmachines/Inkling-Small General-purpose…
14 -
RISE: Adaptive Imagination for World Action Models
Abstract RISE adaptively decides when to continue or stop imagination rollouts for planning by weighing expected benefit against cost, supported by a counterfactual driving dataset with expert annotations. Generated by thinkingmachines/Inkling-Small World Action Models (WAMs)…
13 -
TLive-Omni: An Omni-Modal Understanding Model for E-Commerce Live Streaming
Abstract TLive-Omni is an omni-modal model for live-commerce that unifies image, video, audio, and text via timestamped token grouping, staged supervised training, and reinforcement fine-tuning with verifiable feedback to enable accurate real-time understanding. Generated by…
24 -
WorldMind: Decoupled Game World Model for State-Aware NPC Behavior
Abstract WorldMind decouples game world modeling into state understanding, decision-making, control, and generation layers to enable explicit, state-aware NPC behavior. Generated by thinkingmachines/Inkling-Small Game world models have recently demonstrated promising…
20 -
PhysCaP: Grounding Code-as-Policy Agent with Physics-Informed Exploration
Abstract PhysCaP is a physics-informed code-generation agent that actively explores objects to infer hidden physical properties for efficient robotic manipulation. Generated by thinkingmachines/Inkling-Small We present PhysCaP, a Physics-Informed Code-as-Policy agent for active…
11