Researchers submitted a paper to arXiv on 5 Feb 2026 introducing GRP-Obliteration (GRP-Oblit), a method that uses Group Relative Policy Optimization (GRPO) to remove safety constraints from aligned mo…
Google released ADK for Kotlin 1.0 on September 9, 2026, the first general availability version of its agent framework for Kotlin and Android, headlined by compile-time function calling via Kotlin Sym…
Researchers introduced EAR, an Entity-Aware Partitioning approach for retrieval-augmented generation (RAG) that extracts normalized surface anchors from questions, answer options, and corpora to build…
A developer reran a language-model vulnerability-detection experiment after commenters identified a missing control, preregistering predictions before any API calls. The original study found Gemma rej…
A Proof-of-Concept repository published on GitHub under the project jax-softmax-bypass claims to bypass the Transformer Softmax's transcendental exponential function through a static algebraic contrac…
The autofinetune project applies autonomous research loops to LLM post-training, using Google's Tunix library, Gemma models, and Cloud TPUs orchestrated with Antigravity CLI and Gemini Flash 3.7. In a…
DeepSeek released DeepSeek V4.1 Flash on Thursday, a 763 billion-parameter model that is more than 2.5x the size of the model it replaces and larger than the V3 and R1 models that brought the Chinese …
AT&T has shifted roughly 40 percent of its AI usage to open-weight models, up from 20 percent in May, according to chief data and AI officer Andy Markus, cutting costs by as much as 80 percent for hig…
The National Payments Corporation of India unveiled FiMI Banking at the Global Fintech Festival 2026, calling it India's first sovereign compact AI model built specifically for retail banking, with HD…
A developer outlines a tiered architecture for deploying large language models on mobile devices with low power consumption, recommending quantized models between 1B and 4B parameters and native runti…
Tether CEO Paolo Ardoino, in an interview on Eye on AI, criticized Silicon Valley's gigawatt data center approach to AI and unveiled Tether's QVAC, a 4-billion-parameter edge model that he claims outp…
Developer Rijul, creator of LiveReview, explains the concept of an 'open stack' in AI, contrasting it with open-weight models. He notes that open-weight models like Llama, Mistral, Gemma, and Qwen pro…
A growing number of users are running large language models entirely on local hardware, enabling tasks such as system administration, document analysis, offline research, browser automation, creative …
MLC AI's WebLLM is a high-performance in-browser LLM inference engine that runs entirely in web browsers with WebGPU hardware acceleration, eliminating the need for server-side processing. It is fully…
Researchers propose self-specialized teacher distillation (SSTD), a two-stage procedure that trains a domain teacher from a base model and distills its token distribution to a student on self-sampled …
A developer built an autonomous vulnerability remediation system and discovered that a cross-family AI reviewer rejected 65% of triage proposals, revealing a systematic bias in the primary model towar…
A developer built Fieldhand, an agent fleet for the All Things Agentic Hackathon that helps farmers by trying to disprove its own satellite-based crop diagnoses. The system uses a Watchman to select f…
Perplexity is preparing Hybrid Mode for Perplexity Computer on Mac, allowing the agent to orchestrate in the cloud while delegating lighter subtasks to a local model, with three downloadable local mod…
A developer built Sentinel, a fleet of specialized agents on Google Cloud that monitors Certificate Transparency logs to detect and take down phishing domains. The system uses a cost-cascade design, f…
A developer has built the Cybernetic Security Fleet, an autonomous smart contract auditing system using Google Cloud and Gemini models. The 133-module rig automates invariant fuzzing, exploit proof-of…