AI research papers.

Off-policy data in reinforcement learning with verifiable rewards (RLVR) can arise from asynchronous generation, rollout-training mismatch, and response reuse.

Reinforcement learning (RL) from unit tests is the common recipe for post-training code models, but a unit test only asks whether the final program passes.

Test-time scaling for large language models (LLMs) can involve generating multiple candidate solutions for selection or revising an earlier attempt.

Rubric-based reward models (RMs) evaluate responses against explicit criteria.

System one models enable direct, low-latency decision-making.

A robot needs a world model that predicts how its actions will change the scene and a policy that chooses those actions.

Harness self-evolution revises an agent's software scaffolding, including prompts, memory, workflows, and executable components, while keeping model weights fixed.

With the rapid development of multimodal large language models (MLLMs), evaluating their video understanding capabilities has become increasingly important.

Visual reasoning is an important capability for vision-language models (VLMs) to understand visual evidence, spatial structure, constraints, and state changes.

A central challenge in scaling flow models is maintaining stability in training and integration.

Reconstructing complete, high-quality meshes of large-scale scenes from unposed multi-view images requires faithful geometric grounding while plausibly completing surfaces that are unobserved or only sparsely captured.

Generating information-rich visual artifacts such as slides, posters, and dashboards as code is appealing because the outputs are structured, editable, and verifiable.

How can we build an agent that thinks beyond human language, yet still acts through it in the external environment?

Reasoning models achieve strong performance through long chains of thought (CoT), but often spend more tokens than needed to solve a problem.

Long-horizon agents have drawn growing attention for their ability to work autonomously -- from hours to days -- on economically valuable tasks, a capability essential to automated discovery and recursive self-improvemen…

Sparse-to-dense motion completion recovers dense mesh motion from sparse vertex displacements.

Learning humanoid loco-manipulation from human demonstrations rests on observing how locomotion, posture, and object interaction evolve together as tasks unfold.

Humanoid loco-manipulation requires coordinated whole-body interaction with the physical world, yet collecting robot demonstrations at scale remains costly.

Modern image generation largely relies on specialized generative models that leave the process of image construction implicit in learned parameters.

Sparse attention reduces the KV-cache traffic that dominates long-context LLM decoding, but practical methods largely operate at coarse granularity.

As video models enable longer and more realistic interactive experiences, maintaining a world consistent with its history becomes essential.

Constructing robust GUI agents relies on multimodal visual grounding: the ability to precisely map natural language intents to actionable on-screen components.

Structured pruning reduces LLM inference costs by removing model components to produce smaller dense models.

Vision-language models (VLMs) are increasingly deployed in multi-agent systems to automate complex computer-use workflows, yet existing benchmarks mostly assume consistent, cooperative environments.

How well can a language model turn a large, incomplete and partly contradictory body of documents into a plan when no single correct answer exists?

Can AI agents carry out research on open scientific problems?

Transformers are intuitively considered to be efficient at copying from their inputs.

Agent skills are Markdown documents that a tool-using agent reads as instructions and then acts on with the privileges of the session that loaded them.

Language-model agents sometimes act beyond their authorization and then misreport what they did.

StudentSim is a research system for evaluating and improving AI tutoring tools before they are tested with learners.

Latent diffusion models generate images as codes of an image tokenizer, whose decoder turns them into the pixel image.

Video-based interactive world models aim to enable real-time, user-controllable environments.

LLM-based multi-agent collaboration is increasingly used to solve complex tasks by dividing work among specialized agents and integrating complementary expertise.

Reinforcement Learning (RL) has emerged as a key driver for LLM reasoning.

Finding the right video often requires distinguishing similar scenes in which different events occur.

Reliable deployment of LLM-based multi-agent systems calls for runtime auditors that predict terminal outcomes from observed execution prefixes.

Recent advances in self-improving LLM agents have demonstrated the potential of AI systems that drive their own intelligence.

Reinforcement learning with verifiable rewards (RLVR) typically uses small rollout groups per prompt, where repeated sampling from the same policy can yield highly similar trajectories, reducing within-group reward varia…

Hour-long videos, streaming feeds, and movie-length content are pushing vision-language models (VLMs) past their breaking point -- the standard recipe of encoding each frame into visual tokens yields sequences that grow…

Omni-modal large language models (Omni-LLMs) extend LLMs to process heterogeneous modalities (e.g., vision, audio, and IMU signals) alongside language, enabling a broader range of multimodal tasks.

Recent multimodal large language models (MLLMs) achieve strong results on video question answering, yet often fail to recover the temporal structure that makes a video coherent.

Recent years have witnessed rapid progress in joint audio--visual generation.

On-policy self-distillation (OPSD) trains a large language model using supervision from a teacher instantiated from the same model as the student.

Generalist robot manipulation can benefit from pretrained vision-language models (VLMs) for semantic reasoning and video generation models (VGMs) for world modeling.

Multilingual translation requires a model to preserve meaning while producing text in a requested language.

We study environment scaling as a way to increase agent-benchmark difficulty while holding core task objectives fixed.

Web application development often proceeds through successive requests that extend an existing implementation.

Training LLM agents to interact with users over multiple turns is challenging because the learning signal depends not only on the agent, but also on the user simulator and environment, which jointly shape the agent's exp…