ARC progress update: Competing with sampling — LessWrong

In 2025, the Alignment Research Center (ARC) has been making conceptual and theoretical progress at the fastest pace that I've seen since I first int…

x ARC progress update: Competing with sampling — LessWrong Alignment Research Center (ARC) AI Frontpage 2025 Top Fifty: 11 % 132 ARC progress update: Competing with sampling by Eric Neyman , Victor Lecomte , Wilson Wu , Mikewins , Jacob_Hilton , George Robinson 18th Nov 2025 25 min read 11 132 In 2025, the Alignment Research Center (ARC) has been making conceptual and theoretical progress at the fastest pace that I (Eric) have seen since I first interned in 2022. Most of this progress has come about because of a re-orientation around a more specific goal: outperforming random sampling when it

saved by

• Asher P
• Yudhister Joel Kumar

related reading

• Competing with sampling — Alignment Research Centeralignment.org
• AlgZoo: uninterpreted models with fewer than 1,500 parameters — LessWronglesswrong.com
• Formal verification, heuristic explanations and surprise accounting — Alignment Research Centeralignment.org
• A Mike's-Eye View of ARC's Research — Alignment Research Centeralignment.org
• Fact Finding: Attempting to Reverse-Engineer Factual Recall on the Neuron Level (Post 1) — AI Alignment Forumalignmentforum.org
• Alignment remains a hard, unsolved problem — LessWronglesswrong.com
• Announcing the ARC White-Box Estimation Challenge — LessWronglesswrong.com
• [2605.05179] Estimating the expected output of wide random MLPs more efficiently than samplingarxiv.org
• Obstacles in ARC's agenda: Finding explanations — LessWronglesswrong.com
• Research update: Towards a Law of Iterated Expectations for Heuristic Estimators — Alignment Research Centeralignment.org
• A bird's eye view of ARC's research — Alignment Research Centeralignment.org
• IsoCompute Playbook: Optimally Scaling Sampling Compute for RL Training of LLMscompute-optimal-rl-llm-scaling.github.io