The Archive

Comparison of ML and LLM classifiers for assessing pedagogical quality of AI-generated educational content under distribution shift.

LADA pipeline reduces language annotation bottleneck for vision-language-action models by learning latent action representations from unlabeled trajectories.

Ema has raised $140 million to date and has more than 50 enterprise customers, including Google and Microsoft.

Invideo achieves 3x faster color grading and 50 custom effects/day using GPT-6 Astra for video editing workflows.

Harvey deploys GPT-6 Astra to generate structured, context-aware legal documents with improved drafting precision.

Sam Altman addresses UN Security Council on AI safety, control mechanisms, and need for international governance frameworks.

Category theory framework using groupoids to exploit symmetries in partially observable environments for generalized RL policy learning.

Limiting-Kernel Q(λ) method balances short-horizon n-step and long-horizon value estimation in large-scale RL problems.

MENO neural operator reduces memory footprint for PDE solving with resolution-independent scaling and arbitrary geometric domain support.

Newton-Schulz spectral normalization applied to Vision Transformer attention heads improves training efficiency without adding parameters.

InfiNoVA augments multi-camera robot demonstrations into dense viewpoint distributions via 3D Gaussian reconstruction for viewpoint-invariant VLA policies.

Neural surrogate framework uses fMRI decoding and latent-space steering to predict representational changes for neuromodulation without physical stimulation.

SkillGym framework internalizes human-written agent skills into LLMs via training on 2,756 executable task environments across 12 categories.

FFM-CP fuses multiple pathology vision-language models for few-shot computational pathology classification without retraining.

Proposes validation framework for LLM-generated synthetic survey respondents, addressing intention-behaviour gap in applied decision contexts.

Patreon co-founder Sam Yam is joing OpenAI to lead Creator Product. | Image: The Verge OpenAI has hired three former Patreon execs to anchor its product strategy for creators. After starting the creator subscription platform 13 years ago, co-founder and technology chief Sam Yam announced on X that he's joining OpenAI to lead Creator Product. He's also bringing Patreon's former product head Drew Rowny and engineering head Shannon Ma with him on the new venture, both of whom stepped down in the past few weeks. "We're going to build together with Creators at OpenAI and share early access to a ne...

RefineICL enables tabular foundation models to refine representations in-context via support labels without parameter updates.

Compares nine LLMs and Jev on ContractNLI legal document inference, isolating decision consistency from answer accuracy.

THESEUS framework isolates SLM reasoning capability in knowledge-graph QA by decoupling graph navigation, search, and answer generation.

RARL framework improves RL robustness under distribution shift via risk-sensitive critics and consistency regularization.

Multi-agent RL analysis shows evolutionary stable outcomes may be inaccessible to finite-sample learners in cooperative games.

PRDO algorithm for private decentralized optimization uses recursive estimation and diffusion to handle heterogeneous data and privacy noise.

FLEET adds memory to LLM sampling to reduce duplicate completions and improve aggregation efficiency beyond temperature sampling.

InternW0 from Shanghai AI Lab combines video prediction and robot control via flow matching for actionable physical world modeling.

FedIncome applies federated learning to income estimation for digital lending across state-level data without data pooling.

Survey of brain-to-language decoding covering neural recording, representation learning, and speech/text generation from brain signals.

Traffic forecasting method models local spatial heterogeneity and cross-region context for large-scale prediction.

Stratechery analysis of Meta's Muse platform positioning vs Amazon, Walmart, and Expedia; questions Google's competitive response.

OpenAI releases MentalHealthBench, an expert-informed benchmark for evaluating AI safety and helpfulness in mental health conversations.

Pheno-GS computes geometry-aware optimal transport distances across single-cell patient cohorts at scale.