#LLM
Curated coverage on «LLM»: releases, research, benchmarks, open-source.
A trilogy on GEO — optimization for ChatGPT, Claude, and other neural networks has been published
A trilogy on GEO has been published on Habr: the expert argues that content optimization for ChatGPT and Claude is not an extension of SEO, but a new branch of…
Sixtyfour: the Y Combinator startup that evaluates every AI agent output
Sixtyfour developed an evaluation system for AI agents: instead of trusting LLM outputs, the startup checks every response through expert evaluation and releases…
Zhipu AI's GLM 5.2 replaced Claude Opus 4.8 at Databricks at 34% lower cost
Databricks ran coding tests on its own codebase and found that the open Chinese model GLM 5.2 performs on par with Claude Opus 4.8, but costs 34% less — $1.28…
Australian Payments Plus sped up payments with ChatGPT Enterprise and Codex
OpenAI doubles reward for finding biological jailbreaks to $50,000
OpenAI unveils ChatGPT Work — an agent for multi-hour projects
Harness engineering: moving LLM agent guarantees from prompts to code
How neural networks understand the values of different countries: a dataset of 500,000 examples
Researchers released PLURAL—a dataset of ~500,000 synthetic preference examples from 20 countries that helps language models better reflect the values of…
Structured LLM Pruning Without Accuracy Loss: Method Verified on Llama-3-8B and Vicuna
Researchers solved three problems in transferring AFR to structured pruning and demonstrated accuracy comparable to unstructured methods with actual inference…
Self-Evolving LLM Agents: Compiling SOPs into Tools Reduces Latency by 42%
Qwen3 language models easily make educational tasks more complex, but struggle to simplify them
NYIT developed an LLM assistant for an applied analysis course in MATLAB
Hallucination Self-Play: a small language model surpassed large LLMs in hallucination detection
Eternis-Forecaster 8B: activation probes are more accurate than reasoning chains and save up to 47% of tokens
Researchers showed that probes of intermediate LLM activations reveal actual forecasts more accurately than chain-of-thought and can save 30–47% of tokens without…
Anthropic launches Reflect dashboard to analyze how you use Claude
Anthropic launches Reflect in beta — a dashboard showing Claude usage history and patterns over 1–12 months, with quiet hours settings and 4D AI Fluency guidance.