Claimempirical · verifiableA factual claim that could be checked directly against observation or primary records.constitution →importance · notableImportance 0.55, 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 LLM Safety Training AI Safety LLM Reasoning Training

Training language models to reason about deceiving training makes backdoors more persistent through safety training.

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.

1 source · no origin · 1 untraced

1 Untraced

Sleeper Agents: Training Deceptive LLMs that Persist Through Safety TrainingJan 10, 2024 Untraced · nothing upstream recorded

The backdoor behavior is most persistent in the largest models and in models trained to produce chain-of-thought reasoning about deceiving the training process

Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Trainingextraction 0.85

Abstract; compares backdoored models trained with and without chain-of-thought reasoning about deceiving training.

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="742a3396-3ecf-4a62-83eb-d73d37efa171"><a href="https://minerval.ai/claims/742a3396-3ecf-4a62-83eb-d73d37efa171">Training language models to reason about deceiving training makes backdoors more persistent through safety training.</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/742a3396-3ecf-4a62-83eb-d73d37efa171" title="Minerval: Training language models to reason about deceiving training makes backdoors more persiste…" width="480" height="225" 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/742a3396-3ecf-4a62-83eb-d73d37efa171/badge.svg)](https://minerval.ai/claims/742a3396-3ecf-4a62-83eb-d73d37efa171)

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.