Contribution to the handbook
Supporting · Assessed 2026-10-28
The reported checks using random samples of 5%, 25%, and 50% offer useful context about testing across sample sizes. The packet gives limited results and cautions that they concern the study data, so they do not establish a transferable system test or demonstrated governance capability.
Does this challenge the goal?
In-scope gap
The question concerns variation in sample size during an accuracy check, within the handbook’s existing testing scope; it does not challenge the goal or its assumptions.
The question asks what researchers checked; it is not evidence that these sample sizes are appropriate for governance systems.
Why this category?
- Testing across sample sizes
- Limits on interpreting study-specific results
The supplied passages identify sample sizes and limited results, but do not establish generalizability or suitable sample sizes for other populations or decisions.
This is an AI assessment of contribution to the goal, separate from whether the question has an answer.
Tests, stop conditions and unresolved work →
Authority, accountability and resilience →
Read the exploration goalState whether the researchers made this check and identify the sample sizes used, without inventing results.
What we can say so far
Yes. The researchers compared accuracy using random samples of 5%, 25%, and 50% of participants. The supplied passage does not provide the accuracy values.
What remains open
- The accuracy values are not available in the supplied passage; they are not required to answer whether the check was made or to identify the sample sizes.
Conditions and limits
- The supplied passage does not give the accuracy values or further details about how the random samples were formed.
- The reported sample-size analyses involving 5%, 25%, and 75% concern training-sample results, and the 5%, 25%, and 50% augmentation tests are distinct from the accuracy comparison reported here.
Evidence and related ideas
Supporting evidence, qualifications and context play different roles in this answer.
context
Different training-sample sizes The authors also report results for models trained on random samples of 5%, 25%, and 75% of participants, in addition to the 50% training sample used for the main individual-prediction test.
Describes results reported for models trained on 5%, 25%, and 75% samples; this is a distinct analysis and does not establish the accuracy-comparison sample sizes.
Large Language Models (LLMs) as Agents for Augmented Democracy
Read full source: Large Language Models (LLMs) as Agents for Augmented Democracy (opens in a new tab) ↗Read source discussion: Large Language Models (LLMs) as Agents for Augmented Democracy →
context
Predictions improved some sample matches In the Type I B test, the authors fine-tuned six models on the same preference data: GPT-3.5 Turbo, LLaMA-2 7B, Falcon 7B, Mistral 7B, LLaMA-3 8B and Gemma 7B. They tested initial random samples of 5%, 25% and 50%, adding predictions for 20% of the people who remained outside each sample. For the 5% sample, R² rose from about 30% to about 75%; for the 25% sample, the gain was about 7 to 10 percentage points. The supplied text gives no separate numerical result for the 50% sample.
Describes augmentation tests using initial samples of 5%, 25%, and 50%; this is distinct from the reported accuracy comparison.