#LLM
Curated coverage on «LLM»: releases, research, benchmarks, open-source.
16 Requests, 262 Checks, 6.5 Minutes: Three Autonomous AI Agents Solve One Business Task Live
Sorting sixteen incoming requests from three different channels, cross-checking them against a calendar, a price list, and contracts, takes a live dispatcher two to…
Together AI Launches Post-Training: RL and SFT on B200 at $9 per GPU-Hour
Together AI has opened Post-Training: RL and SFT on dedicated GPUs, LoRA and full weights, and a one-line migration from Tinker. B200 costs $9 per GPU-hour for the…
UniPat AI to Put 100 Million Yuan Into Fellows Program for Scientists
Chinese startup UniPat AI is launching UniPat Fellows: 100 million yuan over one year, 500,000 yuan in API credits per researcher, plus data, experts and compute.
Diffusion LLMs Spend Over Three Times as Many Passes on a Story as on Grammar
The 'Verify Before You Refuse' Rule: How to Stop an AI Agent From Giving Up at the First Error
The Complete Guide to ModelScope: From Model Discovery to Export
Meta Unveils Muse Spark: A Multimodal AI Model With Visual Reasoning
OpenAI Unveils New ChatGPT Voice Powered by GPT Live, With No Technical Details
OpenAI released a video of the updated ChatGPT Voice, built on GPT Live. It gives no latency figures, languages or pricing. Here is what we know and what to test.
Building AI Document-Processing Pipelines with LangExtract and OpenAI
A new guide walks through building pipelines that extract structured data from text using Google's LangExtract and OpenAI language models.
Stateful Harness Lifts World Model Accuracy on Hidden State from 32% to 92%
Hybrid LLMs: Recurrent Layers Suppress Attention, and Staged Training Helps
Halcyon beats LlamaGen and JiT-T2I with 18–275x less pretraining compute
MemT-1.0 builds memory into the model itself: 66x smaller storage at 128K
SEAM teaches TTS new words without audio and cuts data needs by 75%
SEAM initializes embeddings for new TTS tokens from known subtokens, without a single audio recording, and cuts training data requirements by up to four times.
GPT agents lose up to 11.3 points of accuracy when the simulator report sits in a file
When PHREEQC's long output moves from context into a file, GPT-5.1 and GPT-5.2 answer worse by 11.3 and 7.9 points. Claude Opus 4.6 and Sonnet 4.6 show no drop.