We introduce EchoPrompt, a simple yet effective approach that prompts the language model to rephrase its queries before answering them. Just like humans, reminding the model of the question at hand by rephrasing it seems to help preserve context and reinforce connections across tokens. Through extensive experimentation with four causal model families across various tasks—including numerical reasoning (GSM8K, SVAMP, MultiArith, SingleOp), reading comprehension (DROP), and logical reasoning (tracking shuffled objects)—we demonstrate that EchoPrompt introduces substantial improvements in zero-shot and few-shot in-context learning performance without requiring any additional training or external tools. Specifically, EchoPrompt improves Zero-Shot-CoT performance on code-davinci-002 by 5% on numerical reasoning and 13% on reading comprehension tasks on average, and yields consistent gains across GPT-3.5-turbo, Llama-2, and Starcoder.

@inproceedings{mekala2024echoprompt,
  title={EchoPrompt: Instructing the Model to Rephrase Queries for Improved In-Context Learning},
  author={Mekala, Rajasekhar Reddy and Razeghi, Yasaman and Singh, Sameer},
  booktitle={Findings of the Association for Computational Linguistics: NAACL 2024},
  year={2024},
  url={https://arxiv.org/abs/2309.10687}
}

SNOOPY is an online interface for exploring Pile term frequencies alongside precomputed performance of EleutherAI/GPT models on more than 20 NLP benchmarks. It lets users inspect how term frequency relates to few-shot model accuracy across model sizes, example counts, and datasets.

@inproceedings{razeghi-etal-2022-snoopy,
  title = "{SNOOPY}: An Online Interface for Exploring the Effect of Pretraining Term Frequencies on Few-Shot {LM} Performance",
  author = "Razeghi, Yasaman and Mekala, Raja Sekhar Reddy and Logan IV, Robert L and Gardner, Matt and Singh, Sameer",
  booktitle = "Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing: System Demonstrations",
  month = dec,
  year = "2022",
  address = "Abu Dhabi, UAE",
  publisher = "Association for Computational Linguistics",
  url = "https://aclanthology.org/2022.emnlp-demos.39/",
  doi = "10.18653/v1/2022.emnlp-demos.39",
  pages = "389--395"
}

@article{Mumford2020,
  title = {SunPy: A Python package for Solar Physics},
  author = {Mumford, Stuart J. and others},
  journal = {Journal of Open Source Software},
  volume = {5},
  number = {46},
  pages = {1832},
  year = {2020},
  doi = {10.21105/joss.01832}
}

Methods for detecting, classifying, and mitigating noisy-neighbor-induced performance degradation from the tenant side.

@misc{mitra2019predicting,
  title={Tenant-side detection, classification, and mitigation of noisy-neighbor-induced performance degradation},
  author={Mitra, Subrata and Khosla, Sopan and Mehta, Sanket Vaibhav and Mekala, Rajasekhar Reddy and Shah, Aashaka Dhaval},
  year={2019},
  month=nov # "~21",
  note={US Patent Application Publication US20190354388A1, Application No. 15/983,390}
}