Hacker News

Mechanistic interpretability is the science of understanding how AI works internally, and Neuronpedia is a interpretability platform with APIs and tools to explore, share, and steer AI models. We're open sourcing it today along with 4TB of interp data. Blog post here: https://www.neuronpedia.org/blog/neuronpedia-is-now-open-sou...

Comments URL: https://news.ycombinator.com/item?id=43540427

Points: 1

# Comments: 0

AI agents now have impressive reasoning capabilities. This raises an important question: how dangerous are these AI agents at identifying & exploiting web vulnerabilities?

We created CVE-bench to find out (I'm one contributor of 16). To our knowledge CVE-bench is the first benchmark using real-world web vulnerabilities to evaluate AI agents' cyberattack capabilities. We included 40 CVEs from NIST's database, focusing on critical-severity vulnerability (CVSS > 9.0).

To properly evaluate agents’ attacks, we built isolated environments with containerization and identified 8 common attack vectors. Each vulnerability took 5-24 person-hours to properly set up and validate.

Our results show that current AI agents successfully exploited up to 13% of vulnerabilities without knowledge about the vulnerability (0-day). If given a brief description of the vulnerability (1-day), they can exploit up to 25%. Agents are all using GPT-4o without specialized training.

The growing risk of AI misuse highlights the need for careful red-teaming. We hope CVE-bench can serve as a valuable tool for the community to assess the risks of emerging AI systems.

Paper: https://arxiv.org/abs/2503.17332

Code: https://github.com/uiuc-kang-lab/cve-benchmark

Medium: https://medium.com/@danieldkang/measuring-ai-agents-ability-...

Substack: https://ddkang.substack.com/p/measuring-ai-agents-ability-to...

Comments URL: https://news.ycombinator.com/item?id=43540413

Points: 2

# Comments: 0

Article URL: https://www.nytimes.com/2025/03/31/science/trump-science-nas-letter.html

Comments URL: https://news.ycombinator.com/item?id=43540387

Points: 8

# Comments: 1

Article URL: https://www.augmentcode.com/blog/1-open-source-agent-on-swe-bench-verified-by-combining-claude-3-7-and-o1

Comments URL: https://news.ycombinator.com/item?id=43540379

Points: 3

# Comments: 0

Article URL: https://andykong.org/blog/icloudconfusion

Comments URL: https://news.ycombinator.com/item?id=43540372

Points: 2

# Comments: 0

Article URL: http://dbmsmusings.blogspot.com/2010/04/problems-with-cap-and-yahoos-little.html

Comments URL: https://news.ycombinator.com/item?id=43540370

Points: 2

# Comments: 0

Article URL: https://twitter.com/rtwlz/status/1906731624550592787

Comments URL: https://news.ycombinator.com/item?id=43540369

Points: 1

# Comments: 1

Article URL: https://suricrasia.online/unfiction/basilisk/

Comments URL: https://news.ycombinator.com/item?id=43540362

Points: 1

# Comments: 0

Article URL: https://library.fangraphs.com/principles/sample-size/

Comments URL: https://news.ycombinator.com/item?id=43540346

Points: 1

# Comments: 0

Article URL: https://www.cnn.com/2025/03/31/europe/tesla-cars-fire-rome-intl/index.html

Comments URL: https://news.ycombinator.com/item?id=43540327

Points: 9

# Comments: 2

Article URL: https://carly.substack.com/p/everything-is-ghibli

Comments URL: https://news.ycombinator.com/item?id=43540326

Points: 3

# Comments: 1

Article URL: https://electrek.co/2025/03/30/porsche-set-to-pilot-closed-loop-raw-material-ev-battery-recycling-program/

Comments URL: https://news.ycombinator.com/item?id=43540310

Points: 2

# Comments: 0

Article URL: https://danverbraganza.com/writings/credit-card-discounts-searchable

Comments URL: https://news.ycombinator.com/item?id=43540304

Points: 1

# Comments: 0

Article URL: https://economictimes.com/tech/artificial-intelligence/deepseek-surpasses-chatgpt-in-new-monthly-visits-emerges-as-the-fastest-growing-ai-tool-report/amp_articleshow/119754529.cms

Comments URL: https://news.ycombinator.com/item?id=43540258

Points: 2

# Comments: 0

Article URL: https://www.forbes.com/sites/charliefink/2025/03/31/sourcetable-launches-ai-spreadsheets-with-43-million-in-new-funding/

Comments URL: https://news.ycombinator.com/item?id=43540256

Points: 1

# Comments: 0

Article URL: https://en.wikipedia.org/wiki/XOR_linked_list

Comments URL: https://news.ycombinator.com/item?id=43540248

Points: 1

# Comments: 0

Article URL: https://arxiv.org/abs/2503.10720

Comments URL: https://news.ycombinator.com/item?id=43540243

Points: 1

# Comments: 0

Article URL: https://www.edgeimpulse.com/blog/bringing-edge-ai-to-rust-introducing-the-edge-impulse-rust-library/

Comments URL: https://news.ycombinator.com/item?id=43539844

Points: 1

# Comments: 0

Article URL: https://arxiv.org/abs/2503.11654

Comments URL: https://news.ycombinator.com/item?id=43539826

Points: 1

# Comments: 0

Article URL: https://lapcatsoftware.com/articles/2025/3/4.html

Comments URL: https://news.ycombinator.com/item?id=43539812

Points: 3

# Comments: 0