Large language model (LLM) news — GPT-4, Claude, Gemini, Llama, Mistral and the latest research on training, fine-tuning, RLHF, and deployment of LLMs.
A new evaluation framework called PACT is proposed to test whether enterprise-grade LLM agents comply with rules specified in their system context when deployed in sensitive domains such as hiring, healthcare, and financ…
DeepSeek v2 reduces memory footprint by 437x compared with DeepSeek v1, according to a video report. The source provides no further details on the model, its release date, or the benchmark behind the figure.
A developer outlined production techniques for cutting LLM API costs, arguing that prompt-prefix caching, task-based model routing, tighter context management, batch APIs, and hard usage limits deliver far larger savings…
Prompt injection remains an unfixed structural weakness in AI agents because large language models cannot distinguish developer instructions from text they read, and indirect prompt injection — hidden in webpages, emails…
Mozilla added Mistral Small 4 as the fourth AI model in Firefox Smart Window on September 16, 2026, under a zero-data-retention agreement in which Mistral never logs, retains, or trains on user conversations. Smart Windo…
A September 15, 2026 arXiv paper demonstrates that poisoned benchmarks can induce self-modifying AI coding agents to write vulnerable code on clean, held-out tasks, with Hyperagents powered by Sonnet 4.5 self-evolving in…
Anthropic said in a Thursday blog post that Claude now leads 26% of its AI research and development work, up from below 1% in March, with more than 90% of work at or above the "AI collaborates" level. Anthropic reported …
A paper posted to arXiv on 19 May 2026 finds that for large-model pretraining in the high-compute, data-scarce regime, the best data filter is no data filter. The scaling studies report that sufficiently trained large-pa…
A new arXiv paper (arXiv:2609.19213v1) introduces layer-wise curriculum learning for efficient LLM compression, a method that partitions a model into multi-layer segments to transfer knowledge from teacher to student mod…
A new arXiv paper (arXiv:2609.19209v1) proposes an Intent-Driven Query Suggestion Framework that combines intent-aware diversity modeling with query-level credit assignment to generate follow-up query slates. The framewo…
Researchers introduced block parallelism (BP) and context-sharded block parallelism (CSBP), new distributed parallelism dimensions for training block diffusion language models (BDLMs) on long contexts, reporting throughp…
A new technique called probe guidance, introduced in arXiv paper 2609.19356v1, uses the frozen internal states of an existing diffusion model to construct a guidance signal, eliminating the extra forward pass required at…
Researchers developed three methods that use large language models to incorporate rich auxiliary information into Bayesian Optimization, according to arXiv paper 2609.19437v1. The methods consistently outperformed both s…
A new arXiv paper (2609.19499v1) reports that generation schedule, not candidate count alone, determines the energy and latency cost of LLM test-time scaling: on A100 GPUs, eight serial generation calls with one candidat…
A new LSTM Universal Transformer (LSTM-UT) with bounded gated memory outperformed both the Block Universal Transformer (BUT) and CoTFormer on Rule 30 cellular automata and a delayed-recall task, according to an arXiv pap…
Researchers proposed Compressed Active Subspaces (CAS), a method that maps model parameters into a compressed space via a structured isometric embedding before constructing the active subspace, cutting the memory needed …
A systematic mapping of 14,767 arXiv papers introducing or updating LLM evaluation resources between January 2022 and August 2026 finds growing emphasis on action, interaction, and professional applications, with LLM-bas…
A new arXiv study (2609.19244v1) analyzed Web search across four conversational platforms — ChatGPT, Claude, Grok, and DeepSeek — combining real-world user interactions with controlled API experiments. The study found th…
Researchers introduced MAGS, a unified multi-agent framework that generates executable programs with formal safety guarantees by using Dafny as a verification-aware intermediate representation, according to an arXiv pape…
A new arXiv paper (2609.19515v1) introduces LLM-as-an-Improver and the Verify--Repair--Reselect (VRR) method, which uses verification feedback to generate and reselect improved candidates rather than only ranking a fixed…