How funny are the frontier AI models?
This is a crosspost from my substack, where I share results of various experiments in AI and beyond.TLDR: yes, models are getting funnier over timeI love laughing. Well, who doesn’t?
LessWrong9 min read
LessWrong
193 stories
Can you predict cross-propensity spillover?
As in you specifically.tldr: My SPAR group ran a 29 × 29 propensity SFT spillover experiment (i.e. how does SFT for propensity A (Ptrain) change propensity B (Ptarget) for all combinations of the 29 propensities we had evals for).
LessWrong4 min readWhat's the Type of an Ontological Mismatch?
(Brief post, just barely not a quick take. This type signature gets put to work in Rocks to Rules, which locates mismatches on a ladder of binding-types.)Lately I've been thinking about ontological mismatches a lot - even more than I usually might.
LessWrong4 min read
Measuring Chain-of-Thought Necessity and the Effect of Reasoning Constraints
Why readCoT monitoring is a key oversight tool for AI safety, but only if the CoT faithfully reflects the reasoning behind the answer. If it doesn't, monitoring it tells us less about why the model chose that answer.I extend the CoT faithfulness interventions from Lanham et al.
LessWrong8 min read
Encouraging general generalisation: Teaching Gemma (To Articulate) Why
TL;DRBuilding on previous research, I create a constitution for Gemma 3 27B-IT and attempt to align Gemma to it in a way that generalises.I argue that alongside previously considered types of alignment generalisation (across dilemmas and across environments), a new type is worth considering:…
LessWrong29 min read
Takes One to Know One - Training a model to grade reward hacks causes it to reward hack less itself
CodeThe question I asked in this project was - if I finetune a model to judge/catch reward hacks, does it change its behavior when completing the tests itself. Does it learn to hack more or less? Why it’s worth pursuingIt has been shown that narrow finetuning can lead to broad misalignments.
LessWrong8 min read
You can graft SDF changes from base models onto post-trained models
Webpage: praxis-research.org/grafting | Paper link: https://arxiv.org/abs/2610.00767Work done during Peter and Dani's MATS Fellowship (10.0) under the mentorship of Shi, with Jinghua Ou as Research Manager.
LessWrong12 min read
Self-Modeling Interventions Modulate Emergent Misalignment
Code: https://github.com/atagade/sgtr-emModel checkpoints: https://github.com/atagade/sgtr-em/blob/main/MODEL_README.mdFull paper (v2 coming soon): https://arxiv.org/abs/2606.23700TL;DRA model’s self-model is its representation of itself.
LessWrong17 min read
Current AIs out-persuade professionals in lab settings but (probably) not in the world
This article was created by Forethought. See all of our research on our website.How persuasive are present-day AI systems?Unfortunately, it is surprisingly difficult to know where AI capabilities in persuasion are today (September 2026), or even three months ago.
LessWrong21 min read
Anthropic's updated Certificate of Incorporation
As they're preparing for the IPO, I thought I'd retrieve and share Anthropic's updated certificate of incorporation.The PDF contains two documents: the May 27, 2026 certificate and the September 17th amendment to it.(See also the previous discussion.)It is not very interesting.
LessWrong2 min read
Research Note: Filtering Subversion-Relevant Information From Pretraining Data Is Feasible
Authors: Kyle O’Brien¹, Spencer Kitts², Cameron Tice¹, Alek Westover² ¹Geodesic Research, ²Redwood ResearchFigure 1: We pretrained LLMs from scratch with and without filtering subversion-relevant information, finding that subversion-relevant knowledge is significantly diminished while retaining…
LessWrong44 min readDownsides of a Mandatory Training-to-Internal Deployment Gap
I was recently rereading Option 3 from AI Futures’ excellent article How to pace the US frontier: Limit capabilities of models used for AI R&D. In this section they propose a few different options, but their primary proposal is “AI R&D can only be conducted or assisted by AIs that were trained at…
LessWrong8 min read
Reverse Jaggedness and Universal Ajaggedness
Epistemic status: I came up with this an hour ago, this is a fun light-hearted post where I'm enjoying illustrating this idea, which seems true for the most part.[Crossposted from my blog, BlueprintingHeaven.]When it comes to the frontier of AI development, we have this concept of "jaggedness".
LessWrong3 min read
King James’s Dæmonologie, Part I: The Laplacian Mesa-Optimiser
Epistemic status: idle speculation by royal decree of King James VI. In any sufficiently expressive architecture, there exists at initialization a dæmon that is misaligned yet behaves well on every eval, monitor, probe, and training example.
LessWrong1 min read
We Should Build Human-Empowering Software
As AI becomes more integrated into software, there’s a risk that it will exacerbate the flaws of existing tech and disempower humans. It might be useful for some people to work on software that would instead be specifically designed to empower human users and help them achieve their considered…
LessWrong5 min read
Inside the ML4Good Technical AI Safety Bootcamp: What to Expect
TL;DRi) ML4Good is an intensive eight-day technical AI safety bootcamp covering topics such as AI agents, alignment, model reasoning, evaluations, interpretability, and AI governance.ii) It provides useful information about fellowships, career opportunities, starting an AI safety startup, and the…
LessWrong6 min read
Hiring in the AI Safety Ecosystem
Cross-posted from Substack.Below, I’ve included 5 recommendations for hiring managers in AI Safety and 5 tips for candidates as they navigate their career search.
LessWrong6 min read
Reducing synthetic markers makes some SDF false facts linearly indistinguishable from pretraining-acquired knowledge
SummarySynthetic document finetuning (SDF) — the state of the art for implanting false beliefs into LLMs — produces training documents with features that distinguish them from pretraining text.These features (i.e.
LessWrong20 min read
Showing 108 of 193 stories
Load more storiesAiByts Weekly Brief
Get the top AI news and tools straight to your inbox every week.
Weekly · AI news
Join 15,000+ AI enthusiasts