I’m excited to start a new chapter as a PhD Research Scientist Intern at Meta AI, FAIR (Fundamental AI Research) group! Grateful to be part of the CodeGen team in Tel Aviv, working on cutting-edge AI research for code reasoning, understanding and generation 💻🤖

רפי בן שטרית, לשעבר ראש העיר בית שאן, הוא אביו של הלוחם סמ״ר אלרואי ז״ל שנפל בקרב במוצב נחל עוז. אלרואי וחבריו התריעו מראש שחמאס מתכנן מלחמה, אך לא שמעו להם. הקשיבו לדברים מדם ליבו של רפי היקר:

Looking forward to presenting our TACL paper on enhancing LLM creativity at #EMNLP2025 tomorrow (Wed, Nov 5)! 📍 Room A108 🕝 14:30–16:00 (Linguistic Theories, Cognitive Modeling & Psycholinguistics) Details below ⬇️ #NLP #LLMs #Creativity

Heading to #EMNLP2025! 🎉 Two of our papers will be there — come say hi 👋 🖼️ Image Captioning Evaluation — Nov 5, 17:45 📄 https://t.co/TdMVA2iWSD 🕵️ Deceptive LLM Agents (Mafia Game) — Nov 5, 13:00 📄

We present DyPE, a framework for ultra high resolution image generation. DyPE adjusts positional embeddings to evolve dynamically with the spectral progression of diffusion. This lets pre-trained DiTs create images with 16M+ pixels without retraining or extra inference cost. 🧵👇

How can we help LLMs move beyond the obvious toward generating more creative and diverse ideas? In our new TACL paper, we propose a novel approach to enhance LLM creative generation! https://t.co/AFCpQddN6j @ChenShani2 @GabiStanovsky @jurafsky @HyadataLab @stanfordnlp @nlphuji

Excited to share this has now been accepted at #NeurIPS2025 as a position paper (<6% acceptance)!🎉 We advocate for systematically studying entire model populations via weight-space learning, and argue that this requires charting them in a Model Atlas. @NeurIPSConf #NeurIPS 🧵👇

Code World Model: producing code by imagining the effect of executing instructions and planning instructions that produce the desired effect.

We release Code World Model (CWM)! 👩‍💻🌎📊 A coding LLM designed to advance code generation research through agentic reasoning and world-model-based planning. Super excited about this release and proud of the team’s work! 😃 See Gab's post for more info 👇

1/ We released CWM, a 32B dense LLM for coding, agentic use, and, more importantly, to further World-Modeling research. To support this research, we release the pre-training, sft and rl model weights, along with inference code and the tech report. See:

🔥 CWM x BigO(Bench) 🔥 CWM 32B was just released, and evaluated on BigO(Bench) ! Does "world-modeling-aware" training helps CWM reach higher performance on Code Complexity related tasks ?

Our new Code World Model (CWM) is out! I learned and gained expertise working on the RL part, and I'm super proud of what we built. Check out the thread below for the full details.

New from Meta FAIR: Code World Model (CWM), a 32B-parameter research model designed to explore how world models can transform code generation and reasoning about code. We believe in advancing research in world modeling and are sharing CWM under a research license to help empower

new research from Meta FAIR: Code World Model (CWM), a 32B research model we encourage the research community to research this open-weight model! pass@1 evals, for the curious: 65.8 % on SWE-bench Verified 68.6 % on LiveCodeBench 96.6 % on Math-500 76.0 % on AIME 2024 🧵

(🧵) Today, we release Meta Code World Model (CWM), a 32-billion-parameter dense LLM that enables novel research on improving code generation through agentic reasoning and planning with world models. https://t.co/BJSUCh2vtg

🎉 Proud to share that "Debatable Intelligence" has now been accepted to #EMNLP2025 (Main Conference)! https://t.co/zVE73m9lVu Huge thenks to my amazing collaborators @ArielGera2, @RoyBarHaim, @Hoper_Tom, @noamslonim

Proud to share PromptSuite! 🌈 A flexible framework for generating thousands of prompt variations per instance, enabling robust multi-prompt LLM evaluation across diverse tasks. Python API & web UI included. Check it out:

[1/6] 🎬 New paper: Story2Board We guide diffusion models to generate consistent, expressive storyboards--no training needed. By mixing attention-aligned tokens across panels, we reinforce character identity without hurting layout diversity. 🌐 https://t.co/aRG81nu5qK

🚨 Benchmarks tell us which model is better — but not why it fails. For developers, this means tedious, manual error analysis. We're bridging that gap. Meet CLEAR: an open-source tool for actionable error analysis of LLMs. 🧵👇

Presenting my poster : 🕊️ DOVE - A large-scale multi-dimensional predictions dataset towards meaningful LLM evaluation, Monday 18:00 Vienna, #ACL2025 Come chat about LLM evaluation, prompt sensitivity, and our 250M COLLECTION OF MODEL OUTPUTS!