News
AI, cybersecurity, machine learning, scraping & programming news.
699 results on this page · clear filters
Designing Transformers for Continuous Diffusion Language Models
Continuous diffusion language models (CDLMs) offer an alternative approach to parallel language generation; however, most existing studies directly use transformer architectures from either causal large language models (LLMs) or D...
KnowGPO: Knowledge-Grouped Policy Optimization for Search Agents
Deep search agents solve knowledge-intensive tasks through long-horizon multi-turn interactions. Assigning credit to individual search decisions is fundamentally difficult. Outcome-based reinforcement learning provides only a sing...
Training Reasoning Models to Generate Diverse Scientific Ideas
One of the first steps of a research pipeline is generating ideas. Large Language Models LLMs are increasingly used to automate this step, but most of this work optimizes the quality of individual ideas while overlooking diversity...
Learning to Steer LLM Agents and Multi-Agent Systems
Large language model agents and multi-agent systems increasingly tackle complex tasks through long-horizon tool use and inter-agent collaboration. As execution unfolds, these systems are prone to deviating from the overall goal, a...
A tool to compare LLM benchmarks
Developers choice of LLM for a given task often hinges on which model excels in the relevant capability set. Philipp Dubach built Model Atlas with Codex to surface those differences, creating a tool that surfaces how published ben...
LLMs are not AGM (though some have internal monologue)
Leading US LLM vendors consistently frame their roadmaps around achieving artificial general intelligence, positioning themselves as the first to cross that threshold. The narrative suggests that once a model reaches this state, i...
Show HN: Asked LLM to show the most stereotypical man/woman for each country
I asked AI image generators to create the most stereotypical person for every country, running two rounds of the experiment: one in 2023 with Midjourney, and another in 2026 with Grok Imagine, GPT Image, and Gemini. Each run produ...
Show HN: Voice Standard Bench – Open benchmark of LLMs on speech transcripts
AssemblyAI's voice-standard-bench is a Go command-line tool that benchmarks every model behind the llm_gateway across four distinct tasks: summarization, information extraction, text correction, and dictation. The framework runs t...
Tool Policies: one policy engine for every tool your AI agents call
AI agent rollouts frequently stall at the security review stage. A finance team wants an agent to pay supplier bills, but security blocks the request because the payments connector can handle any amount to any supplier and delete...
Trump posts AI-edited photo to brag about White House restoration work
President Donald Trump shared a series of images on Truth Social claiming to show White House renovation progress. One "before and after" image of a column drew scrutiny after Reddit users identified multiple visual inconsistencie...