Research Papers
HarmProfile: New Benchmark Reveals Frontier LLMs' Harmful Outputs Grow with Capability
A new benchmark dataset, HarmProfile, analyzes harmful outputs from 23 frontier LLMs, showing that both harmfulness and diversity...
Prior Labs Unveils Open-Source Suite to Advance Relational Learning
Prior Labs releases RelArena-α, TabPFN-Rel, and RPI, a trio of open-source tools designed to standardize benchmarking, provide a t...
New Benchmark Reveals AI Research Agents' Core Flaw: Lack of Metacognitive Self-Correction
A new study introduces AutoResearchEval, a benchmark of 100 real-world research tasks, and finds that autonomous agents fail prima...
Latent-to-Pixel Training Strategy Boosts Pixel-Space Diffusion Models
A new empirical study from Hugging Face researchers proposes a latent-to-pixel training strategy that accelerates convergence and...
Hugging Face Unveils MOSS-VL: Open Vision-Language Models for Real-Time Interaction
MOSS-VL, a new open vision-language model family, prioritizes real-time interaction by using gated cross-attention to perceive vis...
Hugging Face Unveils Large Discovery Model: A Recurrent AI Engine for Open-Ended Scientific Search
A new recurrent architecture from Hugging Face combines generative modeling with Bayesian non-parametric surrogates to guide uncer...
SA-MRPO: A Saturation-Aware Approach to Multi-Reward RL for Language Models
Researchers introduce Saturation Aware Advantage Reweighting (SA-MRPO), a method that standardizes each reward objective independe...
StateM Runtime Boosts GPT-5.6 to 95.3% Accuracy on Terminal-Bench 2.1
A new runtime system called StateM improves long-horizon agent performance without changing model weights, achieving 95.3% raw acc...
GenRouter: Adaptive Routing Cuts Image-Gen Costs by 95%
Hugging Face researchers introduce GenRouter, a unified routing framework that adaptively assigns prompts to optimal agentic image...
Hugging Face Proposes ACID-Compliant Framework for Reliable LLM Agents
A new research paper from Hugging Face introduces 'agentic transactions,' reinterpreting ACID database guarantees for LLM agents,...
UI-Mate: Open-Weight GUI Agent Sets New Benchmarks with In-Context Demonstrations
Hugging Face researchers introduce UI-Mate, a foundation GUI agent that combines environment-grounded training with in-context dem...
Hugging Face Unveils ClawGym II: Black-Box RL Framework for Agent Harness Optimization
A new black-box reinforcement learning framework from Hugging Face enables stable, scalable optimization of general agents through...