Researcher Shares OpenAI Anthropic Updates and Benchmark Views
Tweet from security researcher discusses recent AI lab announcements and benchmark priorities.
A post by @0x10n, a CMU CSD PhD student focused on security research, highlights material from frontier AI labs. It references separate announcements from OpenAI and Anthropic. The same message connects those items to work on ExploitBench and addresses memorization alongside the viability of open source benchmarks. The author states that benchmarks should still strive to be open source. The visible text is labeled as the first entry in a thread and offers no additional details or external verification.
Combined views
6.3K
2 posts, first seen 22h ago
Reactions
2 postsXI 
Xion@0x10nInteresting updates from frontier AI labs! and https://www.anthropic.com/claude-fable-and-mythos-5-1. With ExploitBench, we've been thinking a lot about memorization and the feasibility of open-source benchmarks. TL;DR: benchmarks should still strive to be open-source. (1/n)
- likes: 85
- replies: 1
- bookmarks: 41
- reposts: 6
XI Xion@0x10nInteresting updates from frontier AI labs! and https://www.anthropic.com/claude-fable-and-mythos-5-1. With ExploitBench, we've been thinking a lot about memorization and the feasibility of open-source benchmarks. TL;DR: benchmarks should still strive to be open-source. (1/n)
- likes: 85
- replies: 1
- bookmarks: 41
- reposts: 6
