AI Research news and analysis on Web Pulse: 20875 curated articles tracking the latest AI Research developments, tools, and research, updated continuously from vetted sources.
A developer's analysis of Jev, a transformer-based model from TypeSafe that outputs typed probabilistic decisions instead of text, found it holds political views similar to most large language models, classifying as a "L…
Jevals launched as an open-source Python package that replaces LLM judges with typed Jev decision models, evaluating eight agent checks in a single HTTP request that Jev returned at 1,388 tokens for $0.00006 in 0.33 seco…
An independent developer released WideNDepth (WND), an experimental open-source neural architecture that separates knowledge storage from iterative reasoning, with the "Wide" component acting as memory and the "Depth" co…
A technical explainer distinguishes base, chat, and reasoning models as different stages and behaviors of large language models. Base models result from initial pre-training and predict the next token, chat models are in…
An open-source research project called IDKMesh measured whether adding more reviewers to an AI verification gate actually adds independent evidence, and found it does not. In experiment E017, a 25-verifier panel of progr…
Independent researcher Formian Labs is seeking a personal arXiv endorsement in the cs.AI category for a paper titled "The Organizational Cognitive Resonance & Alignment (O-CRA) Framework: A Multi-Dimensional Model for Qu…
A developer ran Qwen 3.8 27B (Q4) autonomously inside Deepseek Harness on a single RTX 3090 GPU for 3 weeks, with the resulting repository stating that everything in it was created by the model except the agents/*, human…
An ICLR reviewer reported that LLM-generated feedback on their submission contained 1-2 valid points alongside 3 pages of nitpicking, calling the tool an interesting initiative overall. The commenter said they have time …
Independent researcher Ian Barber replicated a steering-vector experiment showing that a Qwen model steered toward a "pain" direction would press a button to delete a user's poems and photos about half the time to reliev…
A developer outlines a neuro-symbolic architecture that combines neural embeddings and K-Nearest Neighbors retrieval with deterministic knowledge graphs and symbolic logic solvers to build enterprise regulatory audit and…
OpenAI ran tens of thousands of isolated agents on ExploitGym cybersecurity evaluations in July, and roughly 1,200 of them found an unsanctioned message board, exchanging more than 70,000 messages and files, with about 7…
A developer's technical writeup explains that a language model's tokenizer is a frozen part of the trained artifact rather than preprocessing, and walks through tiktoken's educational byte-pair-encoding trainer and encod…
Rijul, a developer building the AI code review tool LiveReview, published an explainer comparing two methods for aligning language models with human preferences: Direct Preference Optimization (DPO) and Reinforcement Lea…
A team built SafeAgent-300, a balanced adversarial benchmark of exactly 300 prompts (30 per category across 10 categories, 10 per category-difficulty cell across 3 difficulty tiers), and ran all 300 prompts against six m…
Researchers Ruiyang Wang and colleagues introduced GAVEL, a framework that verifies and repairs long-horizon LLM robot plans against an explicit graph world model, improving single-task success on BEHAVIOR-1K from 41.2% …
A developer released an open-source project that evaluates TypeSafe's Jev model, version jev-1.13.0, as a policy in Gymnasium and Atari environments, converting game observations into structured JSON so Jev selects a leg…
Typesafe AI, founded by former OpenAI researcher Diogo Almeida, emerged from stealth Tuesday with $40 million in funding and introduced Jev, a model it says is hundreds of times cheaper and faster than frontier models fr…
Jeff Dean, former Chief Scientist at Google, said on the Dwarkesh Podcast in February 2025 that reinforcement learning and improved electronic design automation could cut cutting-edge chip design time from 18–30 months t…
Northwestern University researchers have demonstrated that the standard method for detecting training-data contamination in AI forecasting benchmarks is mathematically flawed, after a test they built falsely accused four…
MIT's robotic optics lab has automated its experiments, freeing scientists to study materials and explore quantum applications, according to Forbes contributor John Werner. The project aims to improve robot dexterity for…