News

Researchers at an undisclosed institution developed a protocol to detect donor-specific functional fingerprints in neural networks after Neural Collapse, using affine-correct alignment to distinguish independently traine…

A proof-of-mechanism study found that a Qwen3.6-27B student model distilled via ontology-amplified techniques matched GPT-5 on grounding Vietnamese financial tasks, but the result was underpowered to establish equivalenc…

Researchers introduced GRID, a grammar-constrained decoding engine for enterprise SQL generation that enforces syntactic validity, role-based access control, and provable guarantees via LALR(1) parser state keys and byte…

Researchers propose a calibration-first reward audit framework for reinforcement learning control in smart greenhouses, enabling comparison of reward components across simulator training, facility-adapted rollouts, logge…

A replay analysis of public LLM agent benchmarks SWE-bench, AppWorld, and tau-bench reveals that the fraction of tasks needed to reach the same pairwise conclusion as the full benchmark varies sharply, from 15% for AppWo…

Researchers introduced PM-Bench, a text-based benchmark to evaluate prospective memory in LLM agents, inspired by cognitive science's Virtual Week paradigm. Testing eight state-of-the-art LLMs across eight configurations…

Researchers introduced CANDI-QA, a new dataset designed to evaluate large language models on context-sensitive question answering in specialized domains like medical diagnostics and financial advisory. Testing over ten m…

Researchers proposed G-SHARE, a guideline-based structured reasoning framework for diagnosing human-factor events in nuclear power plants, which operationalizes the CNNP nine-step diagnostic guideline into a multi-stage …

Researchers evaluated state-of-the-art large language models on Korean-Braille translation and found consistently poor, unstable outputs, revealing a systematic accessibility failure. In contrast, supervised fine-tuning …

Researchers propose a graph-based framework that combines weak supervision with propagation graph analysis to detect disinformation narratives in Telegram ecosystems, modeling how narratives diffuse across interconnected…

Researchers fine-tuned LLaMA 3 (8B) as a drop-in reranker for RAG pipelines, using supervised fine-tuning and 4-bit quantization to replace traditional cross-encoders. The model achieved up to 21% improvement in answer c…

Researchers benchmarked Llama 3.2 for semantic fMRI neural language decoding and improved the Huth encoding-model baseline. They achieved an 11% relative METEOR gain by enhancing the ridge regression pipeline, but found …

A study comparing blind and sighted individuals found that visual experience shapes how people navigate semantic memory. Sighted participants showed higher entropy for abstract concepts, while blind participants had high…

Researchers compared semantic memory retrieval in humans and three large language models (GPT-4o, Gemini-2.5-Pro, Claude-Sonnet-4.5) using verbal fluency data and trajectory-based NLP metrics. Humans exhibited higher ent…

Speechify serves 500,000 dynamic pages to 60 million users on Vercel, achieving 99.99% uptime and a 50% cost reduction after migrating from a hacked infrastructure. The AI work platform uses Next.js and Vercel's caching …

A veteran platform builder with 25 years of experience argues that horizontal platform architectures are fundamentally wrong for the age of AI agents. Citing McKinsey data showing only 6% of organizations see real bottom…

A developer built ai-incident-commander, a CLI tool that automates incident documentation using a local LLM, and tested it against 15 real outages from companies like Cloudflare and GitLab. The tool achieved an 88.3% pas…

A developer open-sourced Menagerai, a self-hosted control plane for managing access, authentication, and usage analytics of vibe-coded internal apps. The tool provides SSO, per-app RBAC, audit logs, and adoption analytic…

A developer migrated 10+ sites from Hostinger ($39/month) to Cloudflare Pages (free), reducing hosting costs by over 95%. The sites were static and didn't need server-side processing. The developer also moved email to Em…

OriginStartAI released an open-source repository with curl, Python, and Node.js examples for making the first OpenAI-compatible API call. The repo covers streaming, JSON structured output, migration notes, and common err…