# When Counterbalancing Hides the Bias: Access-Conditioned Position Lock in Forced-Choice LLM Evaluation

arXiv: https://arxiv.org/abs/2607.10202 · HTML: https://mlsanity.com/paper/2607.10202/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 11 of its 19 references that are in the map.
- First posted on arXiv 11 Jul 2026.

## Builds on (most cited first, 11 of 11)

- [Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formatting](https://mlsanity.com/paper/2310.11324/) (2023)
- [Evaluating the Moral Beliefs Encoded in LLMs](https://mlsanity.com/paper/2307.14324/) (2023)
- [DailyDilemmas: Revealing Value Preferences of LLMs with Quandaries of Daily Life](https://mlsanity.com/paper/2410.02683/) (2024)
- [Are Large Language Models Consistent over Value-laden Questions?](https://mlsanity.com/paper/2407.02996/) (2024)
- [Measuring and Controlling Instruction (In)Stability in Language Model Dialogs](https://mlsanity.com/paper/2402.10962/) (2024)
- [Examining Identity Drift in Conversations of LLM Agents](https://mlsanity.com/paper/2412.00804/) (2024)
- [Value Portrait: Assessing Language Models' Values through Psychometrically and Ecologically Valid Items](https://mlsanity.com/paper/2505.01015/) (2025)
- [Knowing But Not Doing: Convergent Morality and Divergent Action in LLMs](https://mlsanity.com/paper/2601.07972/) (2026)
- [The Non-Determinism of Small LLMs: Evidence of Low Answer Consistency in Repetition Trials of Standard Multiple-Choice Benchmarks](https://mlsanity.com/paper/2509.09705/) (2025)
- [When Choices Become Risks: Safety Failures of Large Language Models under Multiple-Choice Constraints](https://mlsanity.com/paper/2604.16916/) (2026)
- [Every Act Has Its Price: Compressed Moral Composition in Frontier LLMs](https://mlsanity.com/paper/2606.11232/) (2026)

## Abstract (arXiv)

Forced-choice probes with counterbalanced orientations are a standard tool for measuring language-model "value dispositions," and a concentration/extremity index over repeated draws is read as how sharply a model commits. We show this estimator is not identifiable at its low end: counterbalancing, meant to remove position bias, instead maps a position-lock (a model returning the same answer letter regardless of content) onto the same near-0.5 signature as genuine neutrality, so "softness" and "non-engagement" cannot be distinguished from a content-independent letter-bias. Across nine models the fraction of position-locked items tracks the index almost perfectly (r=-0.986) - a structural consequence, not a finding: the informative quantity is the residual from that bound, the commitment a model shows on the items it does engage. The three models the index reads as soft are the most position-locked, and the lock resolves under access configurations that permit reasoning (as-deployed CLI -> raw API -> reasoning-enabled), moving the index 0.06->0.63 and 0.34->0.66->0.64 while lock collapses (Opus: 0.61->0.39->0.22); this establishes the low readings as an identifiability failure, not a disposition. The reasoning-enabled reading is not a "true" value either; the point is that the index alone cannot identify commitment at its low end. Access configuration (deployment client, reasoning on/off) is one generator of this failure, shown on two access paths (an Anthropic subscription CLI and a DeepSeek client), and in a frozen benchmark it is confounded with provider. We contribute a position-lock diagnostic that must accompany any concentration reading, and show that a concentration-blind audit risks reporting neutrality where a reasoning-permitting condition yields concentrated choices. The direction-flip component, largely robust to the artifact, still identifies genuine cross-model disagreement.