Daily AI news, sourcedOn Oct 22, Agent4Sci tests whether coding agents can beat published scientific ML models; Claude Opus improves them by 10.4%.
Where it stands
On Oct 22, Agent4Sci tests whether coding agents can beat published scientific ML models; Claude Opus improves them by 10.4%.
Timeline · newest first
- Official
Agent4Sci tests whether coding agents can beat published scientific ML models; Claude Opus improves them by 10.4%
Posted on arXiv on Oct 22, 2026 (source: arxiv.org).
The paper introduces Agent4Sci, a testbed of 20 scientific ML benchmarks across 11 domains. The authors report frontier agents beat the published best on every benchmark, with Claude Opus at 10.4% mean improvement, mostly through engineering tricks; added research skills raised it to 20.1%. The paper has not yet been peer-reviewed.
Source: arXiv cs.AI
Related stories
Claude37 updates · Models→Anthropic29 updates · Policy & safety→Anthropic IPO8 updates · Business & funding→Anthropic-managed Claude2 updates · Models→
How we label
- Official
- An institution announces something that has already happened, through its own channel or a founder or executive account.
- Statement by …
- A named person's views, predictions, plans or comments, including remarks reported by media when the original is unavailable.
- Reported by …
- Only media or another third party is saying it. We name the outlet.
- Unconfirmed
- Circulating online (posts, comments, leaks). No company, named person or outlet has confirmed it.
Items older than 72 hours never go in Top stories and show their original date.