question
active
question:can-we-use-our-understanding-of-the-character-counting-mechanism-to-construct-a-visual-illusion-for-language-modelsCan we use our understanding of the character counting mechanism to construct a "visual illusion" for language models?
Motivates illusion experiment.
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- Visual illusion.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Character sequences that hijack counting attention heads.
- Language models able to generate images could develop a visual component of their workspace.hypothesis0.769Prediction from verbal-output account.
- Refinement of character-circuit overlap, stressing that self-character is not just another fiction character.
- Antra's earlier definitive statement of the tricameral model.
- Methodological proposal to integrate knowledge from contemplative and cognitive science into AI/artificial life frameworks.
- Paper's assessment of current LLM capabilities relative to Turing Test
- Canonical featurization.
- Articulates why a one-layer transformer with MLP is the appropriate starting target for mechanistic interpretability