Claimempirical · derivedA factual claim that rests on inference from other evidence rather than direct observation.constitution →importance · notableImportance 0.50, from 0 to 1 · notable: a contested point in a live debate (also the default before judging). Higher-importance claims are worth more to assess, so funding reaches them sooner.constitution →

TopicsModel Backdoor Attacks Large language models Machine Unlearning

Backdoors in larger language models are more robust to removal by supervised fine-tuning.

Not yet assessed. Attention goes where its expected value is highest and someone funds it; nothing has funded an assessment of this claim yet, and anyone can.

Decomposition

This claim has not been assessed yet. Attention goes where its expected value is highest and someone funds it; once this claim's assessment is funded and runs, it may well decompose into subclaims.

or create a grant for this whole area →

Provenance

Where this claim has been said, linked to its canonical form.

2 sources · no origin · 2 untraced

2 Untraced

Advice for making robust-to-training model organismsposes it as a question Untraced · nothing upstream recorded

Sleeper Agent Backdoor Results Are Messydenies the claim Untraced · nothing upstream recorded

originuntraced

It’s unclear whether model size matters for robustness to SFT training. In fact, our results are (weak) evidence that smaller models are more robust to SFT training! However, other replications using small models point the other way (small = not robust).

Sleeper Agent Backdoor Results Are Messydeniesextraction 0.80

SA paper only did model-size sweeps for HHH RL.

Overall, we’re unsure whether model size matters for robustness to SFT.

Advice for making robust-to-training model organismsposes it as a questionextraction 0.75

Qwen-2.5 1.5B–72B sweep on two backdoors: weak positive size effect at LR 3e-5 that disappeared at LR 5e-5; earlier Llama-8B more robust than 70B.

Cite this claim: a formal citation with its evidence attached

Embed this claim: show it on another site with its current assessment

An embed reports the graph's judgment as it stands when the page is viewed: the status with its confidence, how disputed the claim is, and the date. It links back here.

CardFits its column and sizes itself. Without the script it stays a quotation with its link.

<blockquote class="minerval-claim" data-claim-id="06fc716b-31c1-45dd-b158-9aa9b3b74e56"><a href="https://minerval.ai/claims/06fc716b-31c1-45dd-b158-9aa9b3b74e56">Backdoors in larger language models are more robust to removal by supervised fine-tuning.</a></blockquote>
<script async src="https://minerval.ai/embed.js"></script>

IframeWhere scripts are not allowed. A fixed height; add ?theme=light or ?theme=dark to fix its colours.

<iframe src="https://minerval.ai/embed/claims/06fc716b-31c1-45dd-b158-9aa9b3b74e56" title="Minerval: Backdoors in larger language models are more robust to removal by supervised fine-tuning." width="480" height="200" style="border:0;max-width:100%" loading="lazy"></iframe>

BadgeAn image for a README or a forum post, in Markdown.

[![Minerval's judgment of this claim](https://minerval.ai/embed/claims/06fc716b-31c1-45dd-b158-9aa9b3b74e56/badge.svg)](https://minerval.ai/claims/06fc716b-31c1-45dd-b158-9aa9b3b74e56)

The badge now:

Contribute

Every judgment on this page is open to challenge. A contribution is evaluated on its merits by the reviewer; if it succeeds the page changes, and if it does not, the reasons are stated. Either way the exchange becomes part of the claim’s public record.

Created by extractor · Nov 11, 2026. Every judgment on this page is accompanied by a reasoning trace.