Daily AI news, sourcedUpdates / Research
Updates Research · Official · 63 items · newest issue first · page 2 of 4
AreaAllModelsAgents & codingApps & productsChips & infraResearchPricing & limitsBusiness & fundingPolicy & safetyOther
- Paper introduces ORL, an on-policy method for learning residual context in diffusion language models
- Paper studies feature elicitation for SAE steering of alignment faking in small LLMsOfficialarXiv cs.LGOct 16Research
- Paper explores three diffusion-LLM designs for generating permutationsOfficialarXiv cs.CLOct 16Research
- Paper compares a non-generative verifier with a frontier LLM for RAG claim verification using confidence-gated cascadingOfficialarXiv cs.CLOct 16Research
- Paper presents MeddeX, a QLoRA-tuned Llama 3 8B RAG system with Gemini Vision for evidence-backed medical report summaries
- Paper proposes PRISM, which uses privileged sibling guidance to post-train generators of multi-agent collaboration topologies
- Paper presents DeepUnfolding Agent, a FastAPI service exposing deep-unfolded adaptive control for classical and quantum ML
- Paper introduces IneqAgent, which combines LLM reasoning with symbolic computation to prove inequalities in Lean 4OfficialarXiv cs.AIOct 16Research
- Paper proposes Counterfactual Neighborhood Risk Fields for evidence-grounded selective safety in vision-language models
- Paper presents AlphaDiana, a measurement system for agents on verifiable reasoning tasks
- Paper proposes value-informed schedule trimming to cut the cost of SMC guidance in discrete diffusionOfficialarXiv cs.AIOct 16Research
- Paper introduces RealtimeBench to test LLM agents in games that keep running while they think
- NVIDIA and IIT Madras' AI4Bharat introduce Nemotron-pretraining-Indic-v1, about 2.8 trillion tokens across 22 Indian languages
- MIT researchers show randomly connected neural networks can self-assemble gradient-style learning from local plasticity rules
- Paper uses deep generative models for phylogenetic inference under complex evolutionary processesOfficialarXiv cs.LGOct 15Research
- Paper proposes loss curvature as a sampling-free proxy for memorization in diffusion modelsOfficialarXiv cs.LGOct 15Research
- Paper studies welfare-aware allocation of scarce human feedback for pluralistic preference alignmentOfficialarXiv cs.LGOct 15Research
- Paper asks whether perturbation predictors learn gene interactions or fail to extrapolate in agentic screeningOfficialarXiv cs.LGOct 15Research
- Paper proposes inverse reasoning models that post-train and verify from answer-only data without gold chain-of-thoughtOfficialarXiv cs.LGOct 15Research
- Study tracks how language models move from copying demonstrations to composing them in in-context learningOfficialarXiv cs.CLOct 15Research