OpenAI releases GPT-5.5, bringing company one step closer to an AI ‘superapp’
OpenAI says its latest model offers increased capabilities across a broad variety of categories.
Search the full wire by company, model, lab, or keyword. Every story we have ever aggregated.
OpenAI says its latest model offers increased capabilities across a broad variety of categories.
Sam Altman highlights GPT-5.5 improvements in token efficiency, latency, and new Codex features.
Anthropic's tightly controlled rollout of Claude Mythos has taken an awkward turn. After spending weeks insisting the AI model is so capable at cybersecurity that it is too dangerous to release publicly, it appears the model fell into the wrong hands anyway. According to…
GPT-5.5 used to generate game logic, suggesting shift toward capability breadth over narrow benchmarks.
GPT-5.5 benchmark results published, showing performance gains across key evaluation suites.
Boris Cherny published post-mortem report on Claude Code issues and system design.
OpenAI releases GPT-5.5 model codenamed 'Spud' per Axios reporting.
OpenAI just announced its new GPT-5.5 model, which the company calls its "smartest and most intuitive to use model yet, and the next step toward a new way of getting work done on a computer." OpenAI just released GPT-5.4 last month, but says that the new GPT-5.5 "excels" at…
Self-supervised learning method detects video speed changes and generates variable-speed content by modeling temporal structure.
Framework shows temporal taskification choices in streaming continual learning affect benchmark conclusions and evaluation stability.
LLMs evaluated as semantic evaluators for ASR, achieving 92-94% human agreement compared to traditional WER metrics.
Analysis demonstrates fine-tuning parameter subspace choice fundamentally alters continual learning problem definition and outcomes.
Theoretical result: multicalibration requires Θ(ε⁻³) sample complexity for achieving Expected Calibration Error bounds.
MathDuels benchmark uses self-play with LLMs as both problem authors and solvers to differentiate model capabilities beyond static benchmarks.
HalluScope benchmark reveals prompt-induced hallucinations in LVLMs stem primarily from language dominance over vision grounding.
Agentic architecture automates conversion of research questions to reproducible scientific workflows via LLM semantic translation and domain expert Skills.
User asks whether local models (32-128GB RAM) deliver genuine productivity or remain hobbyist tools.
Survey frames LoRA through signal processing lens, unifying architectural choices and optimization techniques for parameter-efficient fine-tuning.
Scale-adaptive diffusion framework enables joint spatiotemporal video super-resolution across variable upscaling factors and frame rates.
GiVA uses gradient-informed initialization to improve vector-based adaptation efficiency, matching LoRA training times with extreme parameter efficiency.
Computational framework analyzes Brazilian parliamentary discourse using stylometric analysis and topic modeling on legislative speech.
Nemobot is an interactive environment for creating and deploying LLM-powered game agents across multiple game classes using Claude Shannon's taxonomy.
Study incorporates environmental and visual predictors into motor insurance claim-frequency models using zone-level geographic data.
Multi-stage deep learning framework with warm-start optimization addresses Unit Commitment problem for grid scheduling with renewable integration.
EVENT5Ws is a large manually-annotated open-domain event extraction dataset covering diverse event types for improved algorithm development.
TingIS is an enterprise-scale system for real-time risk event discovery from noisy customer incident data in cloud-native services.
Multimodal text-graph approach leverages LLMs for open-domain event extraction without predefined event types using document-level context.
Claude introduces segmented rate limits: Claude Design separate quota, daily routine runs cap (0/15), and faster reset cycles; suggests potential billing/product changes.