OpenDiscoveryTrace: Process Traces for Evaluating AI Scientist Workflows
arXiv:2609.09203v1 Announce Type: new Abstract: Existing benchmarks for autonomous AI scientists evaluate only final outputs---generated code, hypotheses, or papers---yet discard the reasoning process by which those outputs were obtained. This makes it impossible to audit…
- 01.arXiv:2609.09203v1 Announce Type: new Abstract: Existing benchmarks for autonomous AI scientists evaluate only final outputs---generated code, hypotheses, or papers---yet discard the reasoning process by which those outputs were obtained
- 02.This makes it impossible to audit…
This material is provided as an extended analytical summary.
Read full article at sourceRelated Intelligence
All stories →Claude and Codex Begin Restricting Requests to Hack and Modify Their Own Devices
🚫 Claude and Codex have started limiting requests to hack and modify their own devices. Users complain that AI models from OpenAI and Anthropic are refusing reverse‑engineering tasks — analyzing and altering existing software. One Codex user…
AI Video Without the Mess: From Idea to Five Ready Shots
AI Video Without the Mess: From Idea to Five Ready Shots ⠀ Ask a video model to shoot a bakery video, and you get a mess: the hero changes every frame, the location jumps. Three prompts below turn the idea into five shots where everything holds together. ⠀ 🟠How It Works ⠀ The video — five short…
Anthropic's IPO filing revives demand for AI stocks, lifts US indexes 0.3–0.8%
Revenue of Claude's creator surged 1,304% in six months – to $16.29 billion