When Models Decide and When They Bind: A Two-Stage Computation for Multiple-Choice Question-Answering
Fuente:
arXiv
Guardado en:
| Autores principales: | , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866909983653756928 |
|---|---|
| author | Wong, Hugh Mee Nouwen, Rick Gatt, Albert |
| author_facet | Wong, Hugh Mee Nouwen, Rick Gatt, Albert |
| contents | Multiple-choice question answering (MCQA) is easy to evaluate but adds a meta-task: models must both solve the problem and output the symbol that *represents* the answer, conflating reasoning errors with symbol-binding failures. We study how language models implement MCQA internally using representational analyses (PCA, linear probes) as well as causal interventions. We find that option-boundary (newline) residual states often contain strong linearly decodable signals related to per-option correctness. Winner-identity probing reveals a two-stage progression: the winning *content position* becomes decodable immediately after the final option is processed, while the *output symbol* is represented closer to the answer emission position. Tests under symbol and content permutations support a two-stage mechanism in which models first select a winner in content space and then bind or route that winner to the appropriate symbol to emit. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_03914 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | When Models Decide and When They Bind: A Two-Stage Computation for Multiple-Choice Question-Answering Wong, Hugh Mee Nouwen, Rick Gatt, Albert Computation and Language Multiple-choice question answering (MCQA) is easy to evaluate but adds a meta-task: models must both solve the problem and output the symbol that *represents* the answer, conflating reasoning errors with symbol-binding failures. We study how language models implement MCQA internally using representational analyses (PCA, linear probes) as well as causal interventions. We find that option-boundary (newline) residual states often contain strong linearly decodable signals related to per-option correctness. Winner-identity probing reveals a two-stage progression: the winning *content position* becomes decodable immediately after the final option is processed, while the *output symbol* is represented closer to the answer emission position. Tests under symbol and content permutations support a two-stage mechanism in which models first select a winner in content space and then bind or route that winner to the appropriate symbol to emit. |
| title | When Models Decide and When They Bind: A Two-Stage Computation for Multiple-Choice Question-Answering |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2601.03914 |