Hugging Face daily papers, ranked by community upvotes. Summaries and key terms are written by Workers AI from the abstract — click a term to filter.
This paper develops a new approach to understanding videos by predicting when specific events or evidence occur within the video. Practitioners working on video analysis and AI models might care about this research because it could lead to more accurate and robust video understanding systems.
This paper develops a system to extract and represent skills from human-created resources like videos, code, and articles, allowing software agents to learn from these multimodal inputs. Practitioners might care about this work because it could enable more effective training of agents in various domains.
This paper introduces RAGU, an open-source GraphRAG engine that improves large language models with structured knowledge by separating extraction and consolidation, and trains a compact extractor that outperforms larger models on knowledge-graph construction and GraphRAG tasks. Practitioners might care because RAGU can efficiently generate more accurate and complete context for tasks like factoid-level evidence recall and multi-hop question answering.
This paper introduces a framework called EvolvingWorld that allows characters and worlds to evolve together over time in interactive literary worlds, enabling more realistic and coherent simulations. Practitioners interested in developing more immersive and dynamic interactive stories might care about this approach.
This paper introduces a framework called DeepSearch-Evolve, which helps train self-improving web agents by iteratively refining their performance using their own experience. Practitioners might care because this approach can lead to more efficient and effective agents that can learn from their own mistakes.
This paper proposes a new pruning method for coding agents that prunes tool outputs directly inside the agent, rather than relying on a separate code classifier, and shows it can save up to 39% of tokens while preserving task quality.
This paper introduces a vision-language-action model that can perform mobile manipulation tasks in unseen environments with minimal training data, and how it can be scaled up to achieve better performance. Practitioners might care about this model for building robots that can adapt to new tasks with minimal fine-tuning.
This paper develops a video personalization method that focuses on human-object interactions, aiming to improve the accuracy of video generation by better understanding human-object relationships and incorporating intra-subject references. Practitioners may care about this research as it could lead to more realistic and engaging video content.
This paper introduces a benchmark to evaluate video generation models' ability to reason about physical laws, which is crucial for creating reliable world simulators. Practitioners caring about developing more realistic and physically intelligent AI models will find this research valuable.
This paper introduces RynnBrain 1.1, a family of large-scale embodied foundation models that can perform tasks like spatial reasoning, localization, and planning, and shows promising results in real-world robot experiments. Practitioners may care about the potential of these models for robot manipulation and control.
This paper develops a large audio language model that can answer questions with specific timestamps, improving its ability to understand long audio recordings. Practitioners in audio and speech recognition may care about this development as it enables more accurate and context-specific information retrieval from audio data.
This paper develops a multilingual speech recognition model for underrepresented languages, addressing the issue of data scarcity and uneven performance, and shows promising results in controlled comparisons.
This paper introduces Open-AoE, an open dataset and toolchain for egocentric manipulation learning, providing a scalable and structured platform for training embodied models. Practitioners can use Open-AoE to improve their robot learning models, especially those focused on human-robot interaction and embodied intelligence.
This paper creates a new AI model that can edit and generate videos without needing masks, and can also learn to mimic image editing capabilities. Practitioners might care about this because it could lead to more diverse and realistic video editing data, and enable AI models to understand and generate human-like video editing instructions.
This paper proposes a new type of AI system that can watch and remember things over long periods, and use that memory to reason about the world. Practitioners might care because this could lead to more capable assistants that can learn and adapt over time.
This paper proposes a new method for reinforcement learning in large language models, called Group Entropy-Controlled Policy Optimization (GEPO), which helps balance exploration and exploitation by controlling entropy levels across different tasks. Practitioners might care about GEPO because it can lead to more balanced and task-specific exploration levels.
This paper introduces a new method called REBASE that allows for training-free in-context segmentation, enabling the introduction of new object categories at inference time. Practitioners might care because it eliminates the need for retraining and reduces memory overhead, making it a more efficient approach for real-world applications.
This paper introduces Qwen-Music, a music generation model that can produce highly musical and high-fidelity songs from text descriptions and existing songs. Practitioners in music generation and AI may care about this research for its potential to create realistic and engaging music.
This paper proposes a method to generate synthetic data for training API-calling agents without the need for an actual environment. This is useful for scalability, as collecting high-quality data at scale can be difficult. Practitioners might care about this approach because it could speed up the development of AI agents that can interact with APIs.
This paper investigates the use of the Muon optimizer in reinforcement learning (RL) post-training and finds that it can significantly improve the success rate of RL agents, especially when combined with other techniques like policy optimization and advantage estimation. Practitioners in RL may care about this research to explore new ways to improve the performance of their agents.
This paper evaluates whether language models can be used to generate molecules that fit specific 3D constraints, such as protein pockets and spatial relationships between molecules. Practitioners in the field of drug design might care about this research because it could lead to more efficient and effective methods for generating candidate molecules.
This paper develops a new method for generating high-fidelity 3D images of thin-shell objects, like garments, by learning a continuous surface representation. Practitioners caring about 3D generation and object modeling might care about this approach because it achieves better results with fewer computational resources.
This paper introduces S1-Omni, a unified AI model that can reason about scientific data, generate predictions, and create new scientific content, which could be useful for researchers and practitioners who need to analyze and understand complex scientific information.
This paper proposes a new method for advantage shaping in reinforcement learning called Contrastive Policy Optimization (CPO), which uses contrastive disagreement between reference-guided and vanilla generation distributions to indicate correctness. Practitioners might care because it can improve the effectiveness of reinforcement learning methods in generating correct responses.
This paper develops a new method for generating realistic hand-object interactions in 3D animation, combining appearance and motion to create smooth and believable movements. Practitioners in the field of computer animation and AI may care about this research for its potential to improve the realism and consistency of interactive simulations.
This paper proposes a new way for large language models to learn from feedback, allowing them to retain more detailed information about the quality of their responses and learn from it in a more nuanced way. Practitioners might care because this approach could lead to better performance on tasks where the model doesn't have a clear way to evaluate its own output.
This paper develops a method to transform video foundation models' representations into compact, reconstruction-capable, and generation-friendly video latents, which can be used in various generative modeling tasks. Practitioners can use VideoRAE to improve the performance of their models by leveraging the semantic and spatio-temporal structure captured by the frozen video foundation encoder.
This paper proposes a new method called Distilled Reinforcement Learning that improves large language model post-training by providing fine-grained guidance to transfer new knowledge from a teacher model to a student model. Practitioners might care because it outperforms standard reinforcement learning and on-policy distillation methods in terms of knowledge transfer and model performance.
This paper proposes a new method for training language models to generate coherent and faithful responses, even when the input data has changed significantly. Practitioners might care about this research because it could lead to more robust and adaptable language models that can handle real-world scenarios where data distributions shift.
This paper develops a mathematical framework to analyze the Transformer architecture, using differential geometry to model its core components. Practitioners may care about this work because it provides new insights into the stability and optimization dynamics of Large Language Models.
The post-training of Vision-Language-Action (VLA) models is essential due to the diversity of simulators, robot embodiments, and task objectives. Existing compute services, whether offered as direct accelerator rental or…
Security-agent evaluations commonly measure peak offensive capability under generous inference budgets, emphasizing vulnerability discovery, exploit development, penetration testing, and CTF completion. Such measurements…
Autonomous negotiation agents are increasingly deployed in high-stakes settings such as insurance and procurement. While cryptographic techniques protect explicitly disclosed constraint values, they fail to address a sub…
Vision-language-action (VLA) models predict robot actions from visual observations and language instructions. These actions are defined in the robot's own 3D coordinate frame, yet most VLAs observe the scene in the camer…
Reinforcement learning from verifiable rewards (e.g. GRPO) is the engine behind today's reasoning models, yet it grades only the final answer. On hard problems this trains models to write more rather than to think better…
Recent growth in reinforcement learning (RL) has surfaced a need for diverse, specialized training environments. Hand-curated environments with fixed task and reward difficulties become ineffective signals as model perfo…
Agentic language models must learn when to call tools, when to consume tool responses, and when to answer directly. This makes multi-teacher on-policy distillation a natural training strategy: one teacher can specialize …
Current video generation models achieve impressive results in single-shot generation, yet remain limited in cinematic video generation, where coherent narratives and effective multi-shot composition require explicit shot…
Predicting a football match before kickoff requires more than knowing past results: a model must use changing information and make a clear prediction before the answer is available. We present WorldCupArena, a dynamic be…
Self-hosted AI agents read and write their own memory and configuration files to function. An agent may get compromised via corruption of its own state -- a compromise realized via legitimate OS system call invocation. W…
Real-time multimodal applications, including voice agents and interactive video generation, compose heterogeneous models into pipelines whose efficient deployment requires application-specific decisions about placement, …
The prevailing inference framework for diffusion models formulates generation fundamentally as a problem of numerical integration. This perspective casts the model as an exact estimator, neglecting the inherent statistic…
We introduce Self-Verified Reasoner (SVR-R1), a multi-turn RL framework that turns a model's own verification into a learning signal for multimodal reasoning. For each query, the model proposes an answer using the same w…
Multi-agent systems routinely place one AI agent in authority over another. When a subordinate refuses a task, the manager chooses the outcome: it can renegotiate, report the failure honestly, coerce the subordinate, or …
Large language models (LLMs) have demonstrated growing competence in web page generation. However, existing text-driven approaches rely on complex prompts that impose substantial demands on users and offer limited expres…
Egocentric devices, such as wearable front-facing cameras, provide a unique perspective for capturing the continuous interaction between a human viewer and the surrounding environment. A holistic and efficient multimodal…
Optical coherence tomography (OCT) imaging is essential for the diagnosis and treatment of retinal diseases. Although multimodal large language models (MLLMs) have demonstrated considerable potential in medical image ana…
Scaling robust driving policies is fundamentally bottlenecked by the scarcity of edge cases in curated datasets. While the real world continuously captures these critical events, such long-tail events remain underutilize…
Serial verification gates are a core reliability primitive in LLM harnesses: a candidate answer is returned only if k verifier calls all accept it. Under conditionally independent gates, the recent Odds Law (arXiv:2606.1…
Plasma diagnostic models for tokamak fusion devices are almost universally evaluated on clean, complete sensor data. In practice, fusion diagnostics fail regularly: acquisition systems start late, individual sensors die,…