Loading activity…
@jyangballin
CS PhD @Stanford. Created @SWEbench (multi-lingual/modal); SWE-agent; SWE-smith; InterCode; CodeClash; ProgramBench
AI
Arora shared defense slides titled “Causal Interpretability for Analysing and Improving Language Model Architectures.”
AI
Meta says Muse Wonder launched through its Model API and Muse Code, while Full was available to Muse Code subscribers.
AI
Author Nathan Benaich cites an Anthropic index and argues robotics is nearing a “GPT-2 moment.”
AI
Its creators say the method has an LLM recover a future paper’s idea, charging each hint by how surprising it is.
AI
In an X post, Wenhu Chen congratulated Jiang on graduating and said Jiang had more than 5,000 Google Scholar citations and received a TMLR best paper award in 2025.
AI
Its creators say the benchmark evaluates that ability through 4D inverse graphics.
AI
A post introducing SWE-sweep says top models score below 5% on the benchmark and calls it one of its creators’ most challenging.
AI
A post links to Muse Spark 1.3 evaluation profiles and says trajectories, final codebases and test pass/fail breakdowns have been open-sourced.
AI
A user comparing versions 1.1 and 1.3, both at xhigh, says 1.3 used 39% fewer tokens across the benchmark and brought the number of programs at least 95% resolved from eight to 33.
AI
A user says five-minute Muse Code runs added a language heatmap to each model page and $/task to ProgramBench’s homepage leaderboard.
AI
A post sharing a Microsoft Research paper says Agensh coordinates coding agents through shared state. Agents claim subtasks, share findings, verify results and merge changes asynchronously.
AI
Researchers note its added challenges over earlier benchmark versions and ongoing model gains.
AI
Z.ai released the open model post-trained on its 743B base for coding and agentic tasks.
AI
The AI code editor team announces it has joined SpaceX and will work with SpaceXAI.
AI
Mark Zuckerberg posted that Meta opened weights for a local model with another release planned soon.
