Position Bias Across LLM Model Families
Bibliographic record
Abstract
Large Language Models are being used to solve business problems. Their ability to perform tasks and automate business processes provides value in many industries. Open source models, such as Meta’s Llama models, are becoming popular because they can be used on-premises, which is attractive for regulated industries. They can be embedded as they come in various sizes, from 1 billion to 405 billion parameters. OpenAI This study studies the positional bias of the Meta family of models using sizes of $1,3,8,70$ and 405 billion parameter models. OpenAI has a suite of models that are available on the cloud. We use two multiple-choice question and answer data sets from OpenBookQA and CommonsenseQA. We show that while the larger models exhibit less positional bias than the small models, they nevertheless all exhibit positional bias of some sort. These results have implications for model users to be aware of the bias in the models as they process text with multiple options, as is common with business policies and procedures.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.036 | 0.159 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.002 |
| Bibliometrics | 0.002 | 0.002 |
| Science and technology studies | 0.002 | 0.002 |
| Scholarly communication | 0.005 | 0.007 |
| Open science | 0.003 | 0.003 |
| Research integrity | 0.002 | 0.004 |
| Insufficient payload (model declined to judge) | 0.011 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".