Self-Correcting Models, Smarter Monitors, AI Designs Itself
Three new papers tackle critique dependency in LLMs, ensemble monitoring for AI control, and agents that autonomously discover better neural architectures.
Articles Tagged "LLM"
Best Claude Alternatives in 2026: 7 Models Compared
Seven Claude alternatives compared on API cost, context window, coding performance, and data privacy - from GPT-5.5 and Gemini to open-weight options like Kimi K2.6 and Llama 4.
ChatGPT Gets Bank Access - Day After Data Lawsuit Filed
OpenAI launched ChatGPT Personal Finance on May 15, giving Pro users read access to 12,000+ banks via Plaid - one day after a class action alleged OpenAI shared user conversations with Meta and Google.
arXiv Hits Researchers With 1-Year Ban for AI Slop
ArXiv is issuing one-year submission bans to authors whose papers contain verifiable unvetted AI output, as fabricated academic citations hit a tenfold increase since 2023.
Physics Predicts AI Risk, Math Still Hard, Tokens Saved
A physics formula predicts AI behavioral shifts before they happen, a benchmark shows LLMs fail at 90% of graduate math formalization, and a training-free method cuts synthetic data costs by up to 78%.
SubQ Review: 52x Faster, but Show Your Work
Subquadratic's SubQ claims the first linear-scaling LLM with a 12M-token window - but private beta access, self-reported benchmarks, and a 17-point MRCR gap make independent verification the only test that matters.
Claude Haiku 4.5
Anthropic's fastest and most cost-efficient model, delivering 73.3% on SWE-bench Verified and first-in-family extended thinking and computer use at $1/$5 per million tokens.
Claude 3.7 Sonnet
Anthropic's first hybrid reasoning model with togglable extended thinking, a 200K context window, and state-of-the-art SWE-bench performance at $3/$15 per million tokens.
Olympiad Gold, Broken Memories, and Attention Loss
A 30B model earns IMO gold, memory consolidation silently corrupts agents, and a new metric predicts when LLMs lose track of their instructions.
SubQ
SubQ is the first LLM built on a fully subquadratic attention architecture, achieving a 12M-token research context and 52x faster inference than FlashAttention at 1M tokens.
GPT-OSS 20B
OpenAI's open-weight 21B MoE reasoning model with 131K context, Apache 2.0 license, and o3-mini-level benchmark performance running in 16 GB of memory.
OpenAI o3-pro
OpenAI's maximum-compute reasoning model targets the hardest problems where o3 falls short, at $20/$80 per million tokens.
