Search

What are you looking for?

60 results for rl

AI·Alibaba's Lingbo released six embodied AI models from July 7-10, 2026, introducing 'Embodied Native' — a new paradigm that trains foundation models from scratch on physical-world data, interactions, and causal physics rather than migrating internet LLMs, with LingBot-VA 2.0 as the flagship VLA model achieving real-time 150Hz inference and strong generalization.

36氪 AI·36kr.com··2.5Researchembodied-airoboticsfoundation-model

AI·GeneBench-Pro is a new research-level benchmark with 129 problems across 10 domains in computational biology, testing AI agents on judgment-heavy tasks involving ambiguity, iterative experimentation, and decision-making with synthetic yet realistic datasets.

OpenAI Blog·openai.com··0.2Releasebenchmarkgenomicsbiology

Long-horizon failure in world models is conventionally attributed to compounding error, a generic framing that does not distinguish what kind of error compounds. We propose a kinematic-vs-dynamic reframing: world models tend to imagine kinematically rather than dynamically. We operationalize this as the imagined Kinematic-Consistency Error, a per-step diagnostic that measures how far a rollout departs from a closed-form kinematic null, paired with a perturbation protocol that tests whether iKCE…

HuggingFace Daily Papers·huggingface.co··0.2paper

Introducing LifeSciBench, an expert-authored, expert-reviewed benchmark for evaluating how AI systems handle real-world life science research tasks and decisions.

OpenAI Blog·openai.com··0.2research

Reinforcement learning (RL) for non-verifiable instruction following increasingly relies on LLM judges with prompt-specific rubrics as reward signals. While recent methods adapt these rubrics to the evolving policy during training, the training prompts themselves remain static, drawn from fixed corpora. This static approach often results in a critical misalignment between prompt difficulty and policy capability, leaving the judge unable to recover a discriminative reward signal when prompts fai…

HuggingFace Daily Papers·huggingface.co··0.2paper

Visual policies learned from human videos, teleoperation, and robot demonstrations offer scalable motion priors, but often fail in contact-rich manipulation, where success significantly depends on local force and contact geometry. Tactile sensing provides these complementary signals, yet tactile data remain costly to collect and hard to generalize across sensors, robots, and tasks. We introduce OmniTacTune, a policy-agnostic real-world RL pipeline that adapts tactile feedback to pretrained visu…

HuggingFace Daily Papers·huggingface.co··0.1paper

Pixel-wise Earth-observation (EO) foundation models are now achieving state-of-the-art performance via generated spatial embeddings. However, how these models scale and how best to spend a pretraining budget remain poorly understood. We present the largest controlled scaling study for EO to date: 395 training runs on 1,024 GH200 superchips within a fixed pixel-wise Barlow Twins family, each evaluated on 15 downstream tasks. We find that pretraining loss barely predicts downstream performance (|…

HuggingFace Daily Papers·huggingface.co··0.1paper

The inherent complexity of video understanding makes it difficult to determine whether Video-LLM benchmark performance stems from visual perception, linguistic reasoning, or knowledge priors. While many benchmarks have emerged to assess high-level reasoning, shared criteria for evaluating video understanding remain largely overlooked. Instead of introducing yet another benchmark, we take a step back to re-examine the criteria for evaluating video understanding. In this work, we introduce Video-…

HuggingFace Daily Papers·huggingface.co··0.1paper

Large Audio-Language Models (LALMs) are increasingly integrated into daily applications, yet their generative biases remain underexplored. Existing speech fairness benchmarks rely on synthetic speech and Multiple-Choice Questions (MCQs), both offering a fragmented view of fairness. We propose VIBE, a framework that evaluates generative bias through open-ended tasks such as personalized recommendations, using human-recorded speech. Unlike MCQs, our method allows stereotypical associations to man…

HuggingFace Daily Papers·huggingface.co··0.1paper

Reinforcement learning (RL) has become a central component of post-training large language models (LLMs), yet little is understood about how RL adaptation is distributed across transformer layers. Existing approaches typically update all model parameters uniformly, implicitly assuming that every layer contributes similarly to the gains obtained during RL post-training. In this work, we challenge this assumption through a systematic layer-wise study of RL training. Surprisingly, we find that tra…

HuggingFace Daily Papers·huggingface.co··0.1paper