Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2408.09365 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910839858003968 |
|---|---|
| author | Boateng, Emmanuel Aboah Becker, Cassiano O. Asghar, Nabiha Walia, Kabir Srinivasan, Ashwin Nosakhare, Ehi Srinivasan, Soundar Dibia, Victor |
| author_facet | Boateng, Emmanuel Aboah Becker, Cassiano O. Asghar, Nabiha Walia, Kabir Srinivasan, Ashwin Nosakhare, Ehi Srinivasan, Soundar Dibia, Victor |
| contents | Hand-crafting high quality prompts to optimize the performance of language models is a complicated and labor-intensive process. Furthermore, when migrating to newer, smaller, or weaker models (possibly due to latency or cost gains), prompts need to be updated to re-optimize the task performance. We propose Concept Distillation (CD), an automatic prompt optimization technique for enhancing weaker models on complex tasks. CD involves: (1) collecting mistakes made by weak models with a base prompt (initialization), (2) using a strong model to generate reasons for these mistakes and create rules/concepts for weak models (induction), and (3) filtering these rules based on validation set performance and integrating them into the base prompt (deduction/verification). We evaluated CD on NL2Code and mathematical reasoning tasks, observing significant performance boosts for small and weaker language models. Notably, Mistral-7B's accuracy on Multi-Arith increased by 20%, and Phi-3-mini-3.8B's accuracy on HumanEval rose by 34%. Compared to other automated methods, CD offers an effective, cost-efficient strategy for improving weak models' performance on complex tasks and enables seamless workload migration across different language models without compromising performance. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2408_09365 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Concept Distillation from Strong to Weak Models via Hypotheses-to-Theories Prompting Boateng, Emmanuel Aboah Becker, Cassiano O. Asghar, Nabiha Walia, Kabir Srinivasan, Ashwin Nosakhare, Ehi Srinivasan, Soundar Dibia, Victor Artificial Intelligence Computation and Language Hand-crafting high quality prompts to optimize the performance of language models is a complicated and labor-intensive process. Furthermore, when migrating to newer, smaller, or weaker models (possibly due to latency or cost gains), prompts need to be updated to re-optimize the task performance. We propose Concept Distillation (CD), an automatic prompt optimization technique for enhancing weaker models on complex tasks. CD involves: (1) collecting mistakes made by weak models with a base prompt (initialization), (2) using a strong model to generate reasons for these mistakes and create rules/concepts for weak models (induction), and (3) filtering these rules based on validation set performance and integrating them into the base prompt (deduction/verification). We evaluated CD on NL2Code and mathematical reasoning tasks, observing significant performance boosts for small and weaker language models. Notably, Mistral-7B's accuracy on Multi-Arith increased by 20%, and Phi-3-mini-3.8B's accuracy on HumanEval rose by 34%. Compared to other automated methods, CD offers an effective, cost-efficient strategy for improving weak models' performance on complex tasks and enables seamless workload migration across different language models without compromising performance. |
| title | Concept Distillation from Strong to Weak Models via Hypotheses-to-Theories Prompting |
| topic | Artificial Intelligence Computation and Language |
| url | https://arxiv.org/abs/2408.09365 |