Contribution to the handbook
Supporting · Assessed 2026-10-28
The model and benchmark inventory offers limited technical context that may help interpret studies of AI-assisted preference work. It does not specify a governance system or show that one model type is more suitable, secure, or effective for governance.
Does this challenge the goal?
None identified
The question asks for the compared models and their reported training backgrounds; it does not challenge the handbook’s goal or scope.
The supplied evidence describes model access and training backgrounds but gives no model-by-model comparative results.
Why this category?
- useful technical background
- no demonstrated comparative performance or direct system-design implication
The evidence reports model and training-background distinctions, not how these distinctions affect governance outcomes or resilience.
This is an AI assessment of contribution to the goal, separate from whether the question has an answer.
Tests, stop conditions and unresolved work →
Read the exploration goalNew evidence is being assessed. Any answer below is provisional; parts whose evidence changed or was withdrawn are withheld.
Name the compared models and benchmark, and distinguish their reported training-data and weight-access backgrounds.
What we can say so far
The study fine-tuned six models on volunteer data: LLaMA-2 7B, LLaMA-3 8B, ChatGPT 3.5 Turbo, Mistral 7B, Falcon 7B, and Gemma 7B. It also used BERT SQuAD as a simpler benchmark. The reported backgrounds differ: GPT-3.5 Turbo has proprietary architecture and training data; LLaMA-2, LLaMA-3, Mistral, and Gemma are open-weight but trained on proprietary data; Falcon is open-weight and trained on public data; and BERT is described as a foundation model trained on public data.
What remains open
No material gap is recorded within this part’s current scope.
Conditions and limits
- These are the authors’ reported model and training-background descriptions.
- GPT-4 and Claude 3 were not included because fine-tuning options were not publicly available to the authors at the time.
Evidence and related ideas
Supporting evidence, qualifications and context play different roles in this answer.
supports
Models in the comparison The comparison includes six language models and BERT SQuAD, a simpler benchmark. The models differ in what is known about their training: GPT 3.5 Turbo has proprietary architecture and data; LLaMA-2, LLaMA-3, Mistral, and Gemma are open-weight models trained on proprietary data; Falcon is open-weight and trained on public data; and BERT is a foundation model trained on public data.
Identifies the BERT SQuAD benchmark and reports the models’ weight-access and training-data backgrounds.
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 →
supports
Six fine-tuned models The authors fine-tune six existing language models on the volunteer data: Llama-2 7B, LLaMA-3 8B, ChatGPT 3.5 Turbo, Mistral 7B, Falcon 7B, and Gemma 7B.
Names the six fine-tuned models and gives their reported model 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 →