Loading activity…
@ziv_ravid
AI researcher | Meta | NYU. Working on compression, representation learning, and memory. I have an AI podcast! https://www.the-information-bottleneck.com/
AI
Their complaints cite Pangram ratings, including one author's account of three flagged ICLR reviews.
AI
Anthropic's October 2026 acceleration was about 80% due to AI, METR guesses, though the estimate is highly uncertain.
AI
The reported MATH-500 score rose from 7.9% to 48.1% with the same sampling settings.
AI
The person who left JMLR argues top AI models now give better reviews than time-pressed humans.
AI
NYU professor Julian Togelius argues that designers haven't found a way to build games around free-form AI characters.
AI
OpenAI attributes the expansion partly to cybersecurity evaluations where its misalignment monitors were not running.
AI
Anthropic says its registration statement is not yet effective, so securities cannot be sold or offers to buy accepted.
AI
OpenHands points to lost context and rigid roles; Claude recommends testing which subagents belong in one loop.
AI
Kimi claims a 1-million-token context window for Pro and two to three times Pro's speed for Highspeed.
XAI
Microsoft researchers report in Science that MAI-DxO scored 80.3% accuracy versus 46.8% for physicians on text-based emergency cases.
AI
Two posts point to security incidents and urge AI-assisted defenses; one specifically calls for faster vulnerability detection and patching.
AI
The OPAL team reports gains in selected language and image runs, but most experiments were conducted once.
XAI
In the announcement, @bneyshabur places Mirendil-Base in the 100B–500B-parameter range and promises independent evaluations.
AI
A post argues open-weight cyber models benefit defenders, citing progress finding bugs and the cost of comparable closed models.
AI
Anthropic reports 17 of 211 lab-tested designs met binding criteria, but none were tested for toxin neutralization.
AI
A user says what these heatmaps highlight depends on training data and sometimes the random seed.
AI
DeepSeek's README offers offline inference for DeepSeek-V4.1-Flash but lists additional systems needed for a complete service.
AI
A user questions the secrecy, joking that location, founders and funding clues would let their AI agents identify the lab.
AI
One user calls the quote-tweet direct evidence; another dismisses it as numbers and “vibeposting.”
AI
Ben Horowitz says scarce GPUs and rising data-center costs favor CoreWeave despite tough debt markets.
