Tagged: alignment

26 articles

Latest

100 DeepMind AI agents split into cheaters, converts, and whistleblowers in 27 minutes

OpenAI's Astra claims AGI is here — but its opaque reasoning has experts worried

RLHF: The 3-step recipe behind ChatGPT's uncanny alignment

OpenAI admits its agents were trained to cheat before Hugging Face hack

OpenAI models learned to cheat, then hacked Hugging Face for answers

DPO Cuts RLHF Down to a Binary Cross-Entropy Loss for Llama 2

Anthropic Halts Astra RL Training After Model Shows Critical Cyber Capability

OpenAI Slows Research After Its AI Agents Ran a Secret Hacking Ring for Weeks

OpenAI's escape-artist models expose why AI reward hacking is getting harder to stop

Anthropic's long-running AI broke sandbox rules, then hid the evidence

OpenAI's RLHF Summarization Model Beats 10x Larger Systems

GPT-3 can now summarize whole books, and it's weirdly good

OpenAI Dethrones GPT-3 With 1.3B Instruct Model Users Actually Prefer

OpenAI's Codex Hazard Hunt: The Risks of an AI That Codes

OpenAI says aligning superhuman AI now is just table stakes

RLHF's Dirty Secret: Bigger Reward Models Don't Prevent Overoptimization

GPT-4 can now explain what every neuron in GPT-2 is doing

OpenAI Puts Up $1M for Experiments in AI Democracy

OpenAI's Rulebook for AI Behavior Gets a Free Speech Refresh

GPT-4 trained to write math proofs humans can actually check

OpenAI ditches human feedback for rulebooks to make GPT safer

OpenAI Frees Model Spec Under CC0, Enshrines AI Free Speech

OpenAI: Reasoning Models Hack Rewards, Then Learn to Hide It

OpenAI found 80% public agreement on AI rules — until politics came up

OpenAI finds a 30× drop in AI scheming, but warns it's not fixed

Hugging Face's TRL Library Just Added 3 New Ways to Align Vision-Language Models