Democratizing LLMs: An Exploration of Cost-Performance Trade-offs in Self-Refined Open-Source Models
Open accessAbstract
The dominance of proprietary LLMs has led to restricted access and raised information privacy concerns.High-performing open-source alternatives are crucial for information-sensitive and high-volume applications but often lag behind in performance.To address this gap, we propose (1) A untargeted variant of iterative selfcritique and self-refinement devoid of external influence.(2) A novel ranking metric -Performance, Refinement, and Inference Cost Score (PeRFICS) -to find the optimal model for a given task considering refined performance and cost.Our experiments show that SoTA open source models of varying sizes from 7B -65B, on average, improve 8.2% from their baseline performance.Strikingly, even models with extremely small memory footprints, such as Vicuna-7B, show a 11.74% improvement overall and up to a 25.39% improvement in high-creativity, open ended tasks on the Vicuna benchmark.Vicuna-13B takes it a step further and outperforms ChatGPT postrefinement.This work has profound implications for resource-constrained and informationsensitive environments seeking to leverage LLMs without incurring prohibitive costs, compromising on performance and privacy.The domain-agnostic self-refinement process coupled with our novel ranking metric facilitates informed decision-making in model selection, thereby reducing costs and democratizing access to high-performing language models, as evidenced by case studies.
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Biased self-evaluation directly affects model-based benchmarks: a model's rating can be inflated simply because it is most similar to the model used for evaluation.
“Biased self-evaluation directly affects model-based benchmarks (Shashidhar et al., 2023;Zeng et al., 2023;Yuan et al., 2023;Fu et al., 2023;Li et al., 2024): a model's rating can be inflated simply because it's most similar to the model used for evaluation.”
From LLM Evaluators Recognize and Favor Their Own Generations (2024) · cites this paper as Shashidhar et al., 2023
Section 5 (Discussion; page unavailable) reports that the GPT-4 evaluator favored whichever response appeared first, showing that evaluator bias affected benchmark scores. It does not say that a model's rating was inflated because the model was similar to the evaluator; the bias identified is response-order bias. (LLM, conf=0.86)
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Cites
LoRA: Low-Rank Adaptation of Large Language Models
Hu, E. J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Wang, L., & Chen, W. (2021) · arXiv · doi:10.48550/arxiv.2106.09685 · Cited by 2,523 · see citing works
Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena
Zheng, L., Chiang, W.-L., Sheng, Y., Zhuang, S., Wu, Z., Zhuang, Y., Lin, Z., Li, Z., Li, D., Xing, E. P., et al. (2023) · arXiv · doi:10.48550/arxiv.2306.05685 · Cited by 1,888 · see citing works
Scaling Instruction-Finetuned Language Models
Chung, H. W., Hou, L., Longpre, S., Zoph, B., Tay, Y., Fedus, W., Li, Y., Wang, X., Dehghani, M., Brahma, S., et al. (2022) · arXiv · doi:10.48550/arxiv.2210.11416 · Cited by 1,441 · see citing works
Reflexion: Language Agents with Verbal Reinforcement Learning
Shinn, N., Cassano, F., Berman, E., Gopinath, A., Narasimhan, K., & Yao, S. (2023) · arXiv · doi:10.48550/arxiv.2303.11366 · Cited by 1,087 · cited 6× · see citing works
“While fine-tuning and additional pre-training are effective, they entail additional costs and cannot be adapted to existing models.”
BackgroundMethodInfluentialHellaSwag: Can a Machine Really Finish Your Sentence?
Zellers, R., Holtzman, A., Bisk, Y., Farhadi, A., & Choi, Y. (2019) · Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/p19-1472 · Cited by 830 · see citing works
QLoRA: Efficient Finetuning of Quantized LLMs
Dettmers, T., Pagnoni, A., Holtzman, A., & Zettlemoyer, L. (2023) · arXiv · doi:10.48550/arxiv.2305.14314 · Cited by 507 · see citing works
Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge
Clark, P., Cowhey, I., Etzioni, O., Khot, T., Sabharwal, A., Schoenick, C., & Tafjord, O. (2018) · arXiv · doi:10.48550/arxiv.1803.05457 · Cited by 462 · see citing works
Can Large Language Models Be an Alternative to Human Evaluations?
Chiang, C.-H., & Lee, H.-Y. (2023) · Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · doi:10.18653/v1/2023.acl-long.870 · Cited by 331 · see citing works
Large Language Models Can Self-Improve
Huang, J., Gu, S. S., Hou, L., Wu, Y., Wang, X., Yu, H., & Han, J. (2022) · arXiv · doi:10.48550/arxiv.2210.11610 · Cited by 230 · see citing works
Self-Refine: Iterative Refinement with Self-Feedback
Madaan, A., Tandon, N., Gupta, P., Hallinan, S., Gao, L., Wiegreffe, S., Alon, U., Dziri, N., Prabhumoye, S., Yang, Y., et al. (2023) · arXiv · doi:10.48550/arxiv.2303.17651 · Cited by 219 · see citing works
CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquing
Gou, Z., Shao, Z., Gong, Y., Shen, Y., Yang, Y., Duan, N., & Chen, W. (2023) · arXiv · doi:10.48550/arxiv.2305.11738 · Cited by 57 · cited 2× · see citing works
“(Gou et al., 2023) shows significant improvement in technical tasks with the use of external tools to minimize hallucination and toxicity to provide high quality responses.”
Background
Flagged first, then by citations
Research works printed but not linked: 3
- Elias Frantar, Saleh Ashkboos, Torsten Hoefler, and Dan Alistarh. 2023. OPTQ: Accurate quantization for generative pre-trained transformers. In The Eleventh International Conference on Learning Representa- tions.
- James Kirkpatrick, Razvan Pascanu, Neil Rabinowitz, Joel Veness, Guillaume Desjardins, Andrei A. Rusu, Kieran Milan, John Quan, Tiago Ramalho, Ag- nieszka Grabska-Barwinska, Demis Hassabis, Clau- dia Clopath, Dharshan Kumaran, and Raia Hadsell. 2017. Overcoming catastrophic forgetting in neural networks. Proceedings of the National Academy of Sciences, 114(13):3521-3526. · DOI · Paper page
- Ben Naismith, Phoebe Mulcaire, and Jill Burstein. 2023. Automated evaluation of written discourse coherence using GPT-4. In Proceedings of the 18th Workshop on Innovative Use of NLP for Building Educational Applications (BEA 2023), pages 394-403, Toronto, Canada. Association for Computational Linguistics. · DOI
Cited by
Sorted by citations
LLM Evaluators Recognize and Favor Their Own Generations
Panickssery, A., Bowman, S. R., & Feng, S. (2024) · arXiv · doi:10.48550/arxiv.2404.13076 · Cited by 18 · see citing works
“In methods like reward modeling (Leike et al., 2018; Stiennon et al., 2020), model-based benchmarks (Shashidhar et al., 2023; Zeng et al., 2023; Yuan et al., 2023; Fu et al., 2023; Li et al., 2024), self-refinement (Saunders et al., 2022; Madaan et al., 2023; Lee et al., 2023; Shridhar et al.,…”
- Authors
- Shashidhar, S., Chinta, A., Sahai, V., Wang, Z., & Ji, H.
- Affiliations
- University of Illinois Urbana-Champaign (US)
- Venue
- Findings of the Association for Computational Linguistics: EMNLP 2023
- Published
- 2023
- Pages
- 9070-9084
- OpenAlex
- W4389518896
- Cited by
- 9
- References
- 11
- Citations by year
- 2026: 3 · 2025: 5 · 2024: 1
- FWCI
- 1.59
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- Locations
- Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) (publishedVersion, cc-by)
- arXiv (Cornell University) (submittedVersion, cc-by)
- arXiv (Cornell University) (cc-by)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Keywords
- self-refinement, self-criticism, open-source language models, cost-performance trade-off, model selection, vicuña, ChatGPT, democratization of AI, data privacy, resource-constrained environments
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
