🌏How culturally safe are large vision-language models? 👉LVLMs often miss the mark. We introduce CROSS, a benchmark of 1,284 image-query pairs across 16 countries & 14 languages, revealing how LVLMs violate cultural norms in context. ⚖️ Evaluation via CROSS-EVAL.🧨 Safety

RT @linxins2: Thank you so much Caiming! . We show that involving coding as a new type of action apart from GUI action for CUA can signific….

RT @SFResearch: 🌟 Happy National Intern Day!. Today we celebrate the brilliant minds and diverse perspectives that our interns bring to @SF….

RT @qiancheng1231: 🤝 Can LLM agents really understand us?. We introduce UserBench: a user-centric gym environment for benchmarking how well….

RT @Yihe__Deng: 🙌 We've released the full version of our paper, OpenVLThinker: Complex Vision-Language Reasoning via Iterative SFT-RL Cycle….

RT @alexfabbri4: Excited to share MultiNRC, a new SEAL Leaderboard at Scale AI! MultiNRC is a challenging multilingual reasoning benchmark….

RT @ManlingLi_: Can VLMs build Spatial Mental Models like humans?. Reasoning from limited views?.Reasoning from partial observations?.Reaso….

RT @QiyueGao123: 🤔 Have @OpenAI o3, Gemini 2.5, Claude 3.7 formed an internal world model to understand the physical world, or just align p….

RT @ziqiao_ma: Can we scale 4D pretraining to learn general space-time representations that reconstruct an object from a few views at any t….

RT @victor__li__: Glad to be part of the team!. It's been a great pleasure working with so many talented people at Tesla (both in and out o….

RT @omarsar0: @karpathy Great share as usual! Just read this related piece where a study showed issues with LLM-based agents not recognizin….

RT @tparekh97: 🚨 New work: LLMs still struggle at Event Detection due to poor long-context reasoning and inability to follow task constrain….

RT @yikewang_: LLMs are helpful for scientific research — but will they continuously be helpful?. Introducing 🔍ScienceMeter: current knowle….

RT @steeve__huang: 🚨 The Business AI Plot Thickens 🚨. CRMArena set the stage for business AI evaluation in realistic environments. Now we'r….

RT @StellaLisy: 🤯 We cracked RLVR with. Random Rewards?!.Training Qwen2.5-Math-7B with our Spurious Rewards improved MATH-500 by:.- Rando….

RT @YungSungChuang: 🚨Do passage rerankers really need explicit reasoning?🤔—Maybe Not!. Our findings:.⚖️Standard rerankers outperform those….

Safety data construction by re-purposing the CVQA dataset 🔧

Quantitative comparison of cultural safety performance (English / multilingual) ⚖️

Multi-dimensional categorization of data in CROSS ⬇️

RT @steeve__huang: Cultural safety in AI isn't just nice-to-have, it's essential ✅. Our new paper reveals that leading VLMs struggle with c….

Grateful for the incredible team at UCLA Plus Lab, Salesforce AI Research, and Google DeepMind — @VioletNPeng, Ruichen Zheng, @steeve__huang, and @sunjiao123sun_! 🥳.