Breaking the Chain: A Causal Analysis of LLM Faithfulness to Intermediate Structures
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866915869865541632 |
|---|---|
| author | Somov, Oleg Chaichuk, Mikhail Seleznyov, Mikhail Panchenko, Alexander Tutubalina, Elena |
| author_facet | Somov, Oleg Chaichuk, Mikhail Seleznyov, Mikhail Panchenko, Alexander Tutubalina, Elena |
| contents | Schema-guided reasoning pipelines ask LLMs to produce explicit intermediate structures -- rubrics, checklists, verification queries -- before committing to a final decision. But do these structures causally determine the output, or merely accompany it? We introduce a causal evaluation protocol that makes this directly measurable: by selecting tasks where a deterministic function maps intermediate structures to decisions, every controlled edit implies a unique correct output. Across eight models and three benchmarks, models appear self-consistent with their own intermediate structures but fail to update predictions after intervention in up to 60% of cases -- revealing that apparent faithfulness is fragile once the intermediate structure changes. When derivation of the final decision from the structure is delegated to an external tool, this fragility largely disappears; however, prompts which ask to prioritize the intermediate structure over the original input do not materially close the gap. Overall, intermediate structures in schema-guided pipelines function as influential context rather than stable causal mediators. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_16475 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Breaking the Chain: A Causal Analysis of LLM Faithfulness to Intermediate Structures Somov, Oleg Chaichuk, Mikhail Seleznyov, Mikhail Panchenko, Alexander Tutubalina, Elena Artificial Intelligence Schema-guided reasoning pipelines ask LLMs to produce explicit intermediate structures -- rubrics, checklists, verification queries -- before committing to a final decision. But do these structures causally determine the output, or merely accompany it? We introduce a causal evaluation protocol that makes this directly measurable: by selecting tasks where a deterministic function maps intermediate structures to decisions, every controlled edit implies a unique correct output. Across eight models and three benchmarks, models appear self-consistent with their own intermediate structures but fail to update predictions after intervention in up to 60% of cases -- revealing that apparent faithfulness is fragile once the intermediate structure changes. When derivation of the final decision from the structure is delegated to an external tool, this fragility largely disappears; however, prompts which ask to prioritize the intermediate structure over the original input do not materially close the gap. Overall, intermediate structures in schema-guided pipelines function as influential context rather than stable causal mediators. |
| title | Breaking the Chain: A Causal Analysis of LLM Faithfulness to Intermediate Structures |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2603.16475 |