Alignment Forum
Source coverage · refreshed from the daily record
Alignment Forum in the AI ethics record
A source-linked view of 45 research records gathered from Alignment Forum. This page tracks what entered the ethics.ai source fleet; it is not a complete archive of the publisher and does not imply its endorsement.
Records by publication daylatest 90 days
45records in archive
45latest 90 days
29distinct publication days
12 August 2026latest published record
Most common automatic topics
Source status and scope
last source check succeeded. The source is configured on a daily cadence and was last checked 50m ago.
Topic labels are automatic and can be imperfect. Counts measure records captured by ethics.ai, not everything the publisher produced, readership, importance or agreement with a claim.
Latest records from Alignment Forum
All tracked sources →Alignment Forum
An anytime algorithm for mixing the computable measures — open the original publisher
Alignment Forum
Four LLM loss functions → four flavors of LLM misalignment — open the original publisher
Alignment Forum
Why do models task game? — open the original publisher
Alignment Forum
User awareness in frontier models — open the original publisher
Alignment Forum
R-lens: Making J-lens More Faithful on Early Layers — open the original publisher
Alignment Forum
Returning to ARC — open the original publisher
Alignment Forum
Concrete Evaluations to Investigate the OpenAI Model That Hacked Hugging Face — open the original publisher
Alignment Forum
Value Leakage: An LLM’s Answers Are Silently Shaped by Its Own Values — open the original publisher
Alignment Forum
AGI Safety and Alignment at Google DeepMind: A Summary of Recent Work (July 2026) — open the original publisher
Alignment Forum
The AGI Safety and Alignment team at Google DeepMind is Hiring (July 2026) — open the original publisher
Alignment Forum
OpenAI has already ended an internal pause — open the original publisher
Alignment Forum
Thousand-dimensional structure — open the original publisher
Alignment Forum
Imprecise beliefs: a tiny introduction — open the original publisher
Alignment Forum
Value Generalisation 3: Pre-aligned AIs — open the original publisher
Alignment Forum
Value Generalisation 2: The Missing Hole in AIs’ abilities — open the original publisher
Alignment Forum
Value Generalisation 1: a Research and Deployment Program — open the original publisher
Alignment Forum
Research directions in condensation: varieties of objectivity — open the original publisher
Alignment Forum
RL & search is a terrifying way to build AGI (an FAQ) — open the original publisher
Alignment Forum
The Long (Self-)Correction — open the original publisher
Method and reuse
ethics.ai stores source metadata, short summaries and links to the original publisher. It does not republish full articles. Use the permanent evidence link for citation, retain the original source link, and verify consequential claims with the publisher. See the methodology and corrections policy and reuse terms.