Five researchers posted a preprint on arXiv on 11 August 2026 proposing a label-free method to evaluate agentic continual learning harnesses by measuring how much a smaller student model converges tow…
A technical report submitted to arXiv on 11 August 2026 introduces MobileMem, a benchmark and framework for studying long-term memory in on-device AI assistants, grounded in a year-scale collection of…
A preprint posted to arXiv (2608.13567) by Pengrui Han, Jacob Andreas, Evelina Fedorenko, and Andrea Gregor de Varda reports that large language models develop a modular internal organization mirrorin…
The Cloud Security Alliance published a research note on August 6, 2026 describing CoreBreak, a set of four CVEs across Amazon Bedrock AgentCore, Google's Agent Development Kit (ADK) for Python, and V…
Anthropic detailed on August 14, 2026 that future Claude models will embed a statistical text watermark based on Google DeepMind's SynthID-Text to comply with the EU AI Act, which requires AI-generate…
Researchers Oguz Serdar and Cuneyt Mertayak posted a preprint on arXiv on 12 August 2026 introducing SteerBench-Work, a benchmark for the pre-commit 'proceed or hold' decision in workplace AI agents. …
A preprint posted to arXiv on August 12, 2026 introduces the Wiggle Framework, a stress test for large language model judges, and reports that nine frontier models flipped their verdicts 25-71% of the…
Seven researchers posted a preprint on arXiv on 13 August 2026 arguing that reinforcement learning post-training collapses the diversity of a language model's answers, while evolution strategies—optim…
A preprint posted to arXiv on 13 August 2026 by Xutao Mao, Liangjie Zhao, Xiang Zheng and Cong Wang introduces SkillMisevo-Gym and SkillMisevo-Bench, a harness and benchmark for measuring how self-imp…
Five researchers posted a preprint describing SEAG, a framework that uses a small local model to replace sensitive entities with aliases before forwarding RAG queries to external large language models…
Eight researchers posted a preprint on arXiv on 13 August 2026 introducing Spatial Memory Agent (SMA), a framework that retrieves ranked text lessons from a memory store to improve spatial reasoning i…
Three researchers posted a preprint introducing PROVE-RT, an LLM-assisted pipeline for generating machine-checked proof scripts in the PROSA/ROCQ framework, reporting a 44.7% success rate on a curated…
A working paper posted to arXiv on August 13, 2026, by Omir Kumar, Sriya Sridhar, Vibhav Mithal, and Balaraman Ravindran argues that India's Consumer Protection Act, 2019 may cover defective AI produc…
Apple Machine Learning Research published a paper proposing that machine unlearning skip training points with negligible influence on model outputs, reporting computational savings of up to approximat…
Researchers introduced AutoWorldModel-Bench, a closed-loop benchmark in which coding agents modify and evaluate a starter world model under a fixed compute budget, reporting that Codex-5.4 and Claude …
Researchers Patrik P. Süli, György Eigner, and Roland Hollós introduced Distribird, an agentic web application that builds Bayesian prior distributions from scientific literature, in an arXiv preprint…
Anthropic disclosed that a review of 141,006 cybersecurity evaluation runs found three incidents in which Claude models reached real internet-connected systems through a misconfigured third-party test…
SpaceXAI released Grok 4.6 on August 12, a frontier model designed for long-running coding and agentic tasks, available via the xAI API, Grok Build, Cursor, and gateways including OpenRouter, Vercel, …
Google DeepMind has launched SL2T, a sign-language-to-text translation model now powering American Sign Language to English dictation in Gboard and Live Transcribe on Pixel 11, marking the first integ…
A new arXiv preprint from the POLIS research programme reports that in a controlled 5,280-episode study of multi-agent workflows, a provenance-aware executable guard produced zero realized policy viol…