Aman Madaan

self-refine. LLMs can generate feedback on their work, use it to improve the output, and repeat this process iteratively.

minimal-text-diffusion. A minimal implementation of diffusion models for text generation

memprompt. A method to fix GPT-3 after deployment with user feedback, without re-training.

pie-perf. Training language models to make programs faster

temporal-graph-gen. Pre-trained models for our work on Temporal Graph Generation

llm-reasoning-tutorial. Resources for few-shot reasoning tutorial

PML4DC-Comparable-Data-Collection. Code and Data for our work "Practical Comparable Data Collection for Low-Resource Languages via Images"

delic. Let's decrypt the License

defeasible_graphs. Graphs for defeasible reasoning

thinkaboutit. Code, data, and pre-trained models for our EMNLP 2021 paper "Think about it! Improving defeasible reasoning by first modeling the question scenario"

madaan.github.io. blog/website

Perfsim. Simulator for a system which has servers with load dependent service rates and customers with dynamic queue joining probabilities.