Leon Chen retweeted
🚀 A milestone for OpenRSI: 3 OpenRSI agent-generated optimizations have been merged upstream into SGLang-Omni!
All three PRs underwent rigorous maintainer code review and verification by the SGLang-Omni team.
😉 Real code. Proven improvements. Lasting impact for the open-source community. More to come!
👏 Huge thanks to @zhuofengli96475 and @GenAI_is_real for leading this effort, and to the SGLang team @sgl_project for their support and collaboration!
Try OpenRSI harness now: github.com/OpenRSI-Foundatio…
Generative models of cells mimic microscopy data. But distribution learning optimises at the level of pixels, not biological entities --> so the outputs can appear visually realistic, yet remain biologically unaligned ‼️
We outline a path to post-training for “biological grounding” in virtual cells.
CellFluxRL: Biologically-Constrained Virtual Cell Modeling via Reinforcement Learning
📍#ECCV2026
Leon Chen retweeted
Harbor Adapters and Harbor-Index
arxiv.org/pdf/2609.04298
1+year work, 120+contributors, 300+PRs, 10+ funding partners.
After Terminal-Bench, we’re excited to introduce Frontier Bench — a new coding benchmark built from real-world software engineering tasks contributed by the community.
One thing has become strikingly clear: it’s getting really hard to find coding tasks that frontier AI agents can’t already solve. I can genuinely imagine a future where AI improves faster than we can create new benchmark problems.
frontierbench.ai/
Leon Chen retweeted
We’re releasing Frontier-Bench: a benchmark that measures and evolves with the frontier of agent work.
Built by the team behind Terminal-Bench and Harbor, Frontier-Bench is an on-going community effort.
Frontier-Bench v0.1 contains 74 tasks on which the best agents score ~34%
Leon Chen retweeted
How can we train small agentic models that are highly capable of terminal use and coding?
Announcing OpenThoughts-Agent + OpenThinkerAgent-32B, the strongest Qwen-3 based open-data agentic model: 44.8% avg across 7 agentic benchmarks! (1/n)
J2A brought together researchers, founders, creators, and industry practitioners working across cinematic video generation, world models, controllable creation, editing, VFX, post-production, and creative evaluation.
Huge thanks to our speakers from @runwayml, @GoogleDeepMind, @DecartAI, @moonvalley, @LumaLabsAI, @UtopaiStudios, @pika_labs, and @netflix:
@jimei_yang, @RubenEVillegas, @AbermanKfir, @MateuszOnAI, @wenqi_xian, Jie Yang, @chenlin_meng, @jannekontkanen, @realNingYu
And thank you to our authors, reviewers, attendees, and my co-organizers @felixudr and Stephane Grabli.
From Journey to the Awards to the journey toward Cinematic AI.
See you next year 🚀
cvpr26-j2a.github.io/
#CVPR2026 #J2A #CinematicAI #VideoGeneration #GenerativeAI #WorldModels
The first @Journey2Awards workshop at #CVPR2026 is officially wrapped 🎬
When we proposed @Journey2Awards, the question was simple:
AI video is moving fast — but what does it take to move from impressive clips to movie-grade production?
12
1,091
The first @Journey2Awards workshop at #CVPR2026 is officially wrapped 🎬
When we proposed @Journey2Awards, the question was simple:
AI video is moving fast — but what does it take to move from impressive clips to movie-grade production?
Leon Chen retweeted
Thank you all for making our 1st J2A workshop GREAT! We look forward to hearing from you about the feedback (what do you expect for our next workshop, DM open), and enjoy the rest of @CVPR! See you next time!
organizers in P1: @realleonlc Stéphane Grabli @felixudr
Leon Chen retweeted
Speaking now: Janne Kontkanen @jannekontkanen — Google DeepMind @GoogleDeepMind
"Engineering for the Story: Research and Moviemaking — Then, Now, and Beyond"
Leon Chen retweeted
Speaking now: Wenqi Xian @wenqi_xian — Luma AI @LumaLabsAI
"From Prompting to Directing: Precision and Control in Video Generation"
so excited that 1st @Journey2Awards workshop is happening tomorrow at @CVPR!
We will explore how generative AI and multimodal LLMs can support the full film production pipeline. J2A emphasizes on studio-caliber cinematic video generation, and the integration of generative AI into the complete movie production pipeline, from ideation to post-production.
Leon Chen retweeted
What happens when generative video is built for storytellers, not just clips?
Showcase #2 from @UtopaiStudios explores AI-native cinematic production — created for J2A @CVPR 2026.
Watch the J2A Cinematic AI Showcase 🎬
#CVPR2026 #AICinema
Leon Chen retweeted
From Runway to DeepMind to Netflix and more — the AI video frontier is coming to @CVPR.
J2A features Jie Yang, @MateuszOnAI, @RubenEVillegas, @chenlin_meng, @jimei_yang, @wenqi_xian, @AbermanKfir, @jannekontkanen & @realNingYu.
June 4 · Mile High 4CD
#CVPR2026 #AICinema
Leon Chen retweeted
We have @AbermanKfir from @DecartAI giving a talk at our workshop!
See the schedule at cvpr26-j2a.github.io/




































