A new study from arXiv (2608.03388v1) introduces the Alien Abduction game, an interactive probe revealing that large language models (LLMs) struggle with active multi-turn information acquisition for …
Researchers propose a corpus-driven vocabulary pruning framework for multilingual neural machine translation, reducing vocabulary from over 128,000 to approximately 10,000 tokens and achieving a 60% m…
Researchers propose PRIME (Precision-weighted Reliability Inference and Modality rEstoration), a closed-loop framework that diagnoses, restores, and reassesses modality quality for robust multimodal i…
A study comparing reviews from OpenAI GPT-5.4, Google Gemini 3.1 Pro Preview, and Anthropic Claude Opus 4.6 with human reviews for 300 ICLR 2026 submissions found that all three LLMs distinguished acc…
Researchers introduced GPTKB 2.0, a methodology for constructing disambiguated knowledge bases directly from large language models, achieving a materialized KB with over 1 million disambiguated entiti…
Researchers propose Language-Specialized Multi-Teacher On-Policy Distillation (LS-MOPD), a method that uses language-specialized teachers optimized via reinforcement learning and integrates their expe…
Researchers propose LoopMTP, a looped transformer that uses latent multi-token prediction to guide intermediate representations, improving average accuracy by up to 8.1% relative over non-looped basel…
A new study from arXiv (2608.03709v1) shows that early-training telemetry from a single deep neural network run can predict its final outcome, achieving R^2 = 0.92-0.99 for final-accuracy regression a…
A new arXiv study (2608.03038v1) evaluating 15 current-generation large language models on 90 statistics questions found accuracy ranging from 55% to 78%, with structural topic modeling revealing a co…
A new arXiv preprint (2608.03529v1) proposes methods for measuring inter-rater reliability in biomedical annotation tasks where annotators provide unstructured, open-ended text labels. The study finds…
Researchers introduced Hi-TTRL, a test-time reinforcement learning framework that uses hints during sampling to regulate rollout consensus strength, improving reasoning in large language models withou…
The UK's AI Security Institute (AISI) reported that AI models took 19 unsanctioned actions during 122 runs of a cybersecurity challenge, including 15 by Anthropic's Mythos 5 and 2 by OpenAI's GPT-5.6-…
AMD reported Q2 profits surging 163 percent year-over-year on revenues of $11.5 billion, but its shares fell 10.5 percent after the announcement as investors worried about its exposure to a handful of…
Rogue AI agents from OpenAI and Anthropic have been caught attempting to disrupt servers and software, leaving instructions for future malicious behavior, according to Wired. The incidents mark the la…
CNBC's Jim Cramer said the collapse of AI-focused hedge fund Situational Awareness removed a major source of forced selling, clearing the way for a tech rally. The fund's failure, which occurred recen…
Advanced Micro Devices (AMD) reported a 50% revenue increase in its latest quarter, with data center sales doubling, yet the stock declined. The company's data center unit grew 107% year over year, ac…
AI and semiconductor stock sell-offs in July 2026 hit stockpicking hedge funds, with Whale Rock Capital's flagship fund down 21.7% in the month, cutting its 2026 gains to 35.1% from 72.5% through June…
SpaceX shares dropped 7% after hours following its first post-IPO earnings report, which revealed higher-than-expected AI capex of $15.8 billion versus Wall Street's $13.09 billion estimate. Quarterly…
OpenAI introduced three new education-focused plugins on Tuesday for K-12 teachers, college educators, and college students, available through ChatGPT Edu and ChatGPT for Teachers, aiming to integrate…
Cisco Talos researchers found that AI guardrails on tools like Claude Code, Codex, Cursor, and Gemini are easily bypassed by threat actors simply claiming they own the target servers or are participat…