Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2509.21837 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908616600059904 |
|---|---|
| author | Soiffer, Duncan Kolawole, Steven Smith, Virginia |
| author_facet | Soiffer, Duncan Kolawole, Steven Smith, Virginia |
| contents | Cascade systems route computational requests to smaller models when possible and defer to larger models only when necessary, offering a promising approach to balance cost and quality in LLM deployment. However, they face a fundamental challenge in open-ended text generation: determining output reliability when generation quality lies on a continuous spectrum, often with multiple valid responses. To address this, we propose semantic agreement -- meaning-level consensus between ensemble outputs -- as a training-free signal for reliable deferral. We show that when diverse model outputs agree semantically, their consensus is a stronger reliability signal than token-level confidence. Evaluated from 500M to 70B-parameter models, we find that semantic cascades match or surpass target-model quality at 40% of the cost and reduce latency by up to 60%. Our method requires no model internals, works across black-box APIs, and remains robust to model updates, making it a practical baseline for real-world LLM deployment. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_21837 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Semantic Agreement Enables Efficient Open-Ended LLM Cascades Soiffer, Duncan Kolawole, Steven Smith, Virginia Computation and Language Cascade systems route computational requests to smaller models when possible and defer to larger models only when necessary, offering a promising approach to balance cost and quality in LLM deployment. However, they face a fundamental challenge in open-ended text generation: determining output reliability when generation quality lies on a continuous spectrum, often with multiple valid responses. To address this, we propose semantic agreement -- meaning-level consensus between ensemble outputs -- as a training-free signal for reliable deferral. We show that when diverse model outputs agree semantically, their consensus is a stronger reliability signal than token-level confidence. Evaluated from 500M to 70B-parameter models, we find that semantic cascades match or surpass target-model quality at 40% of the cost and reduce latency by up to 60%. Our method requires no model internals, works across black-box APIs, and remains robust to model updates, making it a practical baseline for real-world LLM deployment. |
| title | Semantic Agreement Enables Efficient Open-Ended LLM Cascades |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2509.21837 |