Is my model perplexed for the right reason? Contrasting LLMs' Benchmark Behavior with Token-Level Perplexity
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917372919545856 |
|---|---|
| author | Prins, Zoë Punzo, Samuele Wildenburg, Frank Cinà, Giovanni Pezzelle, Sandro |
| author_facet | Prins, Zoë Punzo, Samuele Wildenburg, Frank Cinà, Giovanni Pezzelle, Sandro |
| contents | Standard evaluations of Large language models (LLMs) focus on task performance, offering limited insight into whether correct behavior reflects appropriate underlying mechanisms and risking confirmation bias. We introduce a simple, principled interpretability framework based on token-level perplexity to test whether models rely on linguistically relevant cues. By comparing perplexity distributions over minimal sentence pairs differing in one or a few `pivotal' tokens, our method enables precise, hypothesis-driven analysis without relying on unstable feature-attribution techniques. Experiments on controlled linguistic benchmarks with several open-weight LLMs show that, while linguistically important tokens influence model behavior, they never fully explain perplexity shifts, revealing that models rely on heuristics other than the expected linguistic ones. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_29396 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Is my model perplexed for the right reason? Contrasting LLMs' Benchmark Behavior with Token-Level Perplexity Prins, Zoë Punzo, Samuele Wildenburg, Frank Cinà, Giovanni Pezzelle, Sandro Computation and Language Standard evaluations of Large language models (LLMs) focus on task performance, offering limited insight into whether correct behavior reflects appropriate underlying mechanisms and risking confirmation bias. We introduce a simple, principled interpretability framework based on token-level perplexity to test whether models rely on linguistically relevant cues. By comparing perplexity distributions over minimal sentence pairs differing in one or a few `pivotal' tokens, our method enables precise, hypothesis-driven analysis without relying on unstable feature-attribution techniques. Experiments on controlled linguistic benchmarks with several open-weight LLMs show that, while linguistically important tokens influence model behavior, they never fully explain perplexity shifts, revealing that models rely on heuristics other than the expected linguistic ones. |
| title | Is my model perplexed for the right reason? Contrasting LLMs' Benchmark Behavior with Token-Level Perplexity |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2603.29396 |