Loading activity…
@stellalisy
PhD student @uwnlp | visiting researcher @AIatMeta | prev @jhuclsp
AI
Its creators say video reward models struggle with shuffled or held frames, even though humans can perceive the changes.
AI
Its creators say ExploreNet adjusts its exploration distribution to the current state, with rewards tied to rollout diversity.
AI
A post introducing HorizonBench describes a benchmark and data generator for tracking user preferences as life events change them.
AI
A user lists personalization, preference modeling and long-horizon human-agent and multi-agent interactions as research interests.
AI
Two posts list HorizonBench for a lifelong-agent workshop and a multilingual reasoning session for a generative AI workshop.
AI
A prospective attendee lists a chatbot-ads poster at 4:40 p.m. PDT and an October 9 spotlight talk on LLM moral safety.
AI
A post describing a Bloomberg Live clip says InterPositive shot its own dataset on a controlled stage for late-stage training.
AI
A post introducing SWE-sweep says top models score below 5% on the benchmark and calls it one of its creators’ most challenging.
AI
Its creator says the benchmark pairs 10,000 synthetic scene instructions with program verifiers. They report that frontier API models scored at most 22% on a harder challenge set.
