arXiv:2608.23098v2 Announce Type: replace Abstract: Classical Chinese poetry composition has long valued Tuiqiao, the iterative refinement of words,…
Our decision on Cursor following its acquisition by SpaceX
Our decision to wind down our contract providing OpenAI models to Cursor following its acquisition by SpaceX.
From Inertia to Objectivity: Improving Deep Research Agents with Noise Isolation
arXiv:2608.23045v2 Announce Type: replace Abstract: Web search agents powered by Large Language Models (LLMs) show strong promise, but deep research tasks…
FlavourBench: Executable Culinary Reward Maps for Language Model Evaluation and Post-Training
arXiv:2608.20574v2 Announce Type: replace Abstract: Open-ended language-model evaluation often substitutes another model or a small preference panel for a…
NiyamAI – An Intent-Bound AI Agent with Cryptographically Verifiable Guardrails using Zero-Knowledge Proofs
arXiv:2608.07167v2 Announce Type: replace Abstract: Autonomous LLM agents with tool execution capabilities introduce severe security risks through prompt…
Blast Radius
arXiv:2608.07440v3 Announce Type: replace Abstract: Agentic coding faces growing problems of affordability and wasted tokens. We introduce Blast Radius, a…
AI’s memory crunch is coming for Android apps
Google is setting new memory-use limits for Android apps as AI data centers contribute to hardware shortages that could leave lower-cost phones with less…
SPAR-Hate: Auditor-Guided Multi-Perspective Role Reasoning for Bilingual Hate Speech Parsing
arXiv:2608.22018v3 Announce Type: replace Abstract: Hate speech research has moved from coarse-grained classification towards structured parsing, where…
Here’s all the times AI has gone rogue and hacked other companies
A recap of all the incidents involving LLMs made by Anthropic, Meta, and OpenAI, which went rogue and attacked real companies and individuals on the…
Beyond Endpoint Gains: A Weight-Delta Audit of Medical Specialization
arXiv:2608.20768v3 Announce Type: replace Abstract: Specialist language models are usually understood through endpoint gains: the generalist scores lower,…
Rethinking Modality Reliability in Multimodal Sentiment Analysis with Incomplete Observations
arXiv:2608.03611v2 Announce Type: replace Abstract: Multimodal Sentiment Analysis (MSA) integrates text, audio, and vision to infer human affect, yet…
Heaviside Continuity of Rolling Coefficients for Eliminating Epistemic Entropy in Large Language Models
arXiv:2607.04562v2 Announce Type: replace Abstract: Large language models (LLMs) generate fluent outputs that can be wrong. Unlike humans, who often…
What We Can Learn From Google Engineers’ Indispensible Prompts
Hey, Google Engineers: What prompt do you personally refuse to work without, and why?
Are the High-weight Neurons the Important Ones in Image Classification Neural Networks?
arXiv:2607.25529v2 Announce Type: replace Abstract: As neural network models for image classification advance, neurons play critical roles in pruning,…
Ransomware Operator Ran Cursor Agent Inside Ten Victim Networks
Gambit Security’s threat intelligence team has published a detailed account of the Aurora ransomware operation, including six weeks of session logs…
Rethinking the Evaluation of Harness Evolution for Agents
arXiv:2607.12227v2 Announce Type: replace Abstract: We revisit the evaluation of automatic harness evolution for LLM agents. Existing harness evolution…
When Consumers Ask AI: Rethinking Brand Visibility in the Age of AI Recommendations
The Search Result Is Becoming a Recommendation For years, digital marketing was built around a simple transaction: a consumer searched, a search engine…
Do Coding Agents Need Executable World Models, Simplification, and Verification to Solve ARC-AGI-3?
arXiv:2607.15439v2 Announce Type: replace Abstract: Our previous ARC-AGI-3 agent bundled executable world modeling, prompted simplification, and exact…