The Archive

Fork-Think: decide-first-then-reason paradigm for LLM inference, identifying confident forking points to avoid overgeneration in parallel sampling.

Primal-dual optimization framework for constrained online convex optimization without Slater regularity assumptions.

SAGE: autonomous research agent with multi-hypothesis failure attribution for robust experiment recovery beyond single-reflection feedback loops.

TabPATE: differentially private PATE-style defense for tabular in-context learning without public data, defeats membership inference.

Von Mises and evidential deep learning for uncertainty quantification in automotive radar direction-of-arrival estimation.

CLOUDADV uses zero-shot LLM forecasting to recommend cloud VM instance sizes under workload drift, reducing overprovisioning.

Team MKC applies LLMs to detect mental health changes from social media timelines for early detection and monitoring.

CSTrader is a multi-agent LLM framework for language-grounded trading in niche Counter-Strike 2 weapon skin markets.

GoodQ uses off-the-shelf generative models for zero-shot quantization-aware training of edge-deployed object detectors.

UniTac is the first unified multimodal model for tactile understanding and generation across cross-sensor configurations.

Algorithms for contextual slate bandit problems under batched and rarely-switching limited-adaptivity regimes with GLM rewards.

Study revises RVL-CDIP dataset by fixing label errors and test-train overlap, re-benchmarking document classification models.

Epistemological study on affective sovereignty and who determines emotion meaning in embedded emotion-sensing AI systems.

CDR-Bench evaluates LLM execution of compositional, order-sensitive data refinement recipes across 3,462 tasks in four domains.

BiRG-LoRA is a rank-gated LoRA adapter with input-conditioned dimensions for parameter-efficient medical question answering.

OKX is bringing together payments, identity and reputation into a marketplace for AI agents.

OpenAI publishes adoption metrics showing ChatGPT user growth and expanded capability exploration across regions.

Commentary on a slow news day in AI; no substantive developments or announcements.

Google UK publishes economic impact report on AI adoption and productivity benefits, positioning open-weights models as tools for broader UK workforce enablement.

A new report finds "high-intensity AI adopters” saw headcount increase 10.2%. Among those companies, entry-level headcount rose by 12%, countering the rhetoric that AI kills junior jobs.

Wix-owned vibe coding platform Base44 has started rolling out its own AI model — with hopes that it will eventually outperform frontier models.

OpenAI debugged rare infrastructure crashes using core dump analysis, revealing 18-year-old software bug and hardware fault.

OpenAI releases GeneBench-Pro, a benchmark for evaluating AI on genomics and biology tasks using real-world datasets.

Technical deep-dive into GeneBench-Pro benchmark design, methodology, and evaluation framework for biology AI models.

Simon Willison releases HTML table extractor tool supporting conversion to Markdown, CSV, JSON, and other formats.

South Korea targets physical AI lead and commercial humanoid robots by 2028.

Google is expanding Gemini’s personalized AI image generation to eligible free users in the U.S., allowing the chatbot to create images based on your interests and data from connected Google apps.

Tidal shared its new policies regarding AI-generated music today and how the platform plans to "protect artists" and "inform listeners." Instead of banning it outright, starting on July 15th Tidal will label tracks it has identified as being 100 percent AI-generated with an icon. But starting today those tracks will no longer be monetizable. "Tidal's priority is ensuring royalties go to original works directly produced, written, and performed by people. We will therefore not knowingly attribute royalties to music we identify as wholly AI-generated," the company's announcement reads. The platf...

OpenAI is releasing some sort of device related to its AI-powered coding tool, Codex, on July 15th. In a video posted to X on Monday, OpenAI shows a square-shaped device with several buttons, alongside the caption, "Your favorite Codex shortcuts are getting an upgrade." This isn't the mysterious AI-powered device OpenAI is working on with former Apple designer Jony Ive, however. As shown in the teaser, OpenAI is launching the device in partnership with Work Louder, a company that sells an array of mechanical keyboards and macro pads with mappable keys, dials, and switches. The silhouette of t...