Daily AI news, sourcedUpdates / Research
Updates Research · Official · 63 items · newest issue first · page 3 of 4
AreaAllModelsAgents & codingApps & productsChips & infraResearchPricing & limitsBusiness & fundingPolicy & safetyOther
- Paper uses pretrained language models as zero-shot tools for record linkage and name-based demographic inferenceOfficialarXiv cs.CLOct 15Research
- Paper argues LLMs can use pretrained world knowledge for event prediction when the task is framed differentlyOfficialarXiv cs.CLOct 15Research
- Study benchmarks eight LLMs on zero-shot abstractive summarization for English and Bengali XL-SumOfficialarXiv cs.CLOct 15Research
- Paper trains voice agents to follow legally required call flows when transferring to unseen debt-collection proceduresOfficialarXiv cs.CLOct 15Research
- Researchers use Gemini 4 agents to find new exact plasma equilibria that break stellarator symmetry, extending counterexamples to Grad's 1967 conjectureOfficialarXivOct 15Research
- NVIDIA researchers' AlphaNIST trains one robot assembly policy from binary rewards alone to 98.2% success on all 20 NIST board tasks in simulationOfficialarXiv (NVIDIA)Oct 15Research
- Google uses Gemini 4 Argon to turn videos of people playing cornhole into robot training simulationsOfficialX @GoogleAIOct 14Research
- OpenAI studies the internal signals behind metagaming in an o3 reinforcement learning runOfficialOpenAI Alignment Research BlogOct 14Research
- Looped-DiT paper shows a 260M looped text-to-image model beating a 6.5× larger baseline with less inference compute
- Researchers release U-Space, a label-free way to read LLM uncertainty from hidden states in a single pass
- Google DeepMind's AlphaProof Nexus paper, which solved 9 open Erdos problems in Lean, is published in Science
- Microsoft researchers' Fork-and-Flush clears agent chat history but keeps the workspace, lifting long-run resultsOfficialarXiv (Microsoft Research)Oct 13Research
- Qwen introduces QwenGyre, an elastic RL framework for training agents on extremely long tasks
- BrokenML benchmark tests how much cluster access AI agents need to fix broken ML jobs
- CL4SM opens a call for contributions on constrained learning for scientific modelling ahead of NeurIPS Paris
- Sakana AI publishes MASS, where one model organizes copies of itself into research teams and trains on what worked
- Microsoft tests how reliably Foundry's multi-turn evaluators grade voice-agent calls
- Prima Mente launches the Alzheimer's Translation Challenge, an AI competition built on a 150M-cell atlas
- Google's AMIE research chatbot pre-screened 98 urgent care patients with no safety stops, per a Lancet studyOfficialGoogleOct 9Research
- Anthropic: Claude Science helps build first complete ultraviolet all-sky mapOfficialAnthropicOct 9Research