# Vision-Language Asymmetry in Bistable Image Captioning

arXiv: https://arxiv.org/abs/2606.08031 · HTML: https://mlsanity.com/paper/2606.08031/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in VLM benchmarks and visual reasoning, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 7 of its 14 references that are in the map.
- First posted on arXiv 6 Jun 2026.

## Placed in

- Multimodal LLMs / vision-language models / [VLM benchmarks and visual reasoning](https://mlsanity.com/mllm/vlm-benchmarks-and-visual-reasoning/)

## Builds on (most cited first, 7 of 7)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Improved Baselines with Visual Instruction Tuning](https://mlsanity.com/paper/2310.03744/) (2023)
- [Sparse Autoencoders Learn Monosemantic Features in Vision-Language Models](https://mlsanity.com/paper/2504.02821/) (2025)
- [Steering CLIP's vision transformer with sparse autoencoders](https://mlsanity.com/paper/2504.08729/) (2025)
- [A Philosophical Introduction to Language Models - Part II: The Way Forward](https://mlsanity.com/paper/2405.03207/) (2024)
- [Evaluating Vision-Language Models on Bistable Images](https://mlsanity.com/paper/2405.19423/) (2024)
- [Mechanistic Interpretability Needs Philosophy](https://mlsanity.com/paper/2506.18852/) (2025)

## Abstract (arXiv)

Wittgenstein's duck-rabbit poses a question for vision-language models: when a model captions an ambiguous image, where in the model is the commitment to one aspect made? We address this with a 3,320-generation behavioral baseline over 83 bistable stimuli that surfaces three regimes (default-dominant, force-dominant, force-balanced) under neutral vs forced-choice prompting, then probe the underlying representations using a TopK sparse autoencoder we train on the CLIP layer that LLaVA-1.6-7B actually consumes (validation EV 0.93). Across 69 bistable stimuli with both per-aspect feature pools available, 72% (50/69) show simultaneous activation of both pools at the vision tower, including 12/12 default-dominant duck/rabbit and 7/8 force-balanced young/old. Causal steering at CLIP layer 22 flips captions on default-dominant stimuli (33% rabbit-flip rate under a fluency guard) but cannot flip captions on force-balanced young/old at any tested coefficient, despite their vision-side superposition. The dominance bottleneck lives downstream of the vision tower; the gap between vision-side representation and language-side commitment is an empirical handle on the seeing/seeing-as distinction. We also flag a methodological note: rank-based statistics on TopK SAE outputs require tie-corrected ranking to avoid silent row-order bias.