Minerval

← the claim it bears on

A line of reasoning from premises to a conclusion. Like a claim, it belongs to no one who makes it.constitution →

The argument from benign fine-tunes that also improve under noise

This argument, if it holds, weighs against the claim.constitution →Increasing Gaussian weight noise improves benchmark performance in sandbagging LLMs, unlike the degradation in non-sandbagging models.

The argument

The claim states a contrast: sandbagging models gain under weight noise and non-sandbagging models only degrade. When sandbagging was installed by fine-tuning a 70B model to imitate a weaker one, the gains under noise were small (maximum scores mostly 1.0 to 1.4 times the unperturbed score), one organism showed no extra gain on the task it was built to sandbag, and the same models given their password, which were not sandbagging, also gained. Noise raises scores wherever a model is under-elicited and cannot tell benign under-elicitation from sandbagging. A second trained organism showed no gain on one of its two benchmarks. The contrast therefore does not hold for sandbagging models in general.

Does it work?

Does it follow?
If the premises are true, the conclusion is much more likely, though not guaranteed.constitution →
Do the premises hold?
1 supported

Granting its premises, the general statement fails: a contrast that disappears for trained sandbaggers measured against fine-tuned comparators is not a property of sandbagging models as such. It does not show the effect is absent everywhere, and it leaves the prompted case untouched. The weight rests on the finding that noise lifts under-elicited models whether or not they are sandbagging, which so far comes from one unblinded post hoc analysis of two 70B organisms with thinly reported methods and has not yet been assessed in its own right. The one other null on a trained lock with a described procedure adds a little to the first half of the argument; a larger set of reported zeros on eleven password locks adds almost nothing, because its noise procedure is undescribed and its figures are inconsistent with a full sweep.

Premises

Where it is made

Passages where sources make, criticize or mention this argument. They record where it appears, not how strong it is, and none has precedence over another.

  • Auditing Games for Sandbaggingmakes itThe passage states the inference with what it needs.constitution →
    The fact that password-unlocked and password-locked models show the same trend suggests that fuzzing detects general underelicitation, rather than specifically detecting strategic sandbagging -- models show uplift proportional to how underelicited they are relative to the base model, regardless of whether that underelicitation is intentional or benign.

    Taylor et al. · 2025-12

Earlier ratings

  • Nov 14, 2026If the premises are true, the conclusion is much more likely, though not guaranteed.constitution →

    Granting its premises, the general statement fails: a contrast that disappears for trained sandbaggers measured against fine-tuned comparators is not a property of sandbagging models as such. It does not show the effect is absent everywhere, and it leaves the prompted case untouched. The weight rests on the finding that noise lifts under-elicited models whether or not they are sandbagging, which so far comes from one unblinded post hoc analysis of two 70B organisms with thinly reported methods and has not yet been assessed in its own right. The null results on other trained locks support the first half of the argument independently of it.