Summary of Paramanu-ayn: Pretrain From Scratch or Continual Pretraining Of Llms For Legal Domain Adaptation?, by Mitodru Niyogi et al.
PARAMANU-AYN: Pretrain from scratch or Continual Pretraining of LLMs for Legal Domain Adaptation?
by Mitodru Niyogi, Arnab Bhattacharya
First submitted to arxiv on: 20 Mar 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper presents Paramanu-Ayn, a collection of legal language models trained exclusively on Indian legal case documents. The 97-million-parameter Auto-Regressive (AR) decoder-only model was pretrained from scratch with a context size of 8192 on a single GPU for just 185 hours, achieving an efficient MFU of 41.35. The authors also developed a legal domain specialized BPE tokenizer. The model’s performance was evaluated using perplexity and zero-shot tasks: case judgment prediction with explanation and abstractive case summarization. Paramanu-Ayn outperformed Llama-2 7B and Gemini-Pro in the case judgment prediction task, despite being 72 times smaller. In zero-shot abstractive summarization, it surpassed decoder-only LLMs generating fixed-length summaries (5000 tokens) by over 10 percentage points in BLEU and METEOR metrics, and by nearly 4 percentage points in BERTScore. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper introduces a new language model called Paramanu-Ayn. It’s special because it was trained on Indian legal documents and can do things like predict case judgments and summarize cases. The model is smaller than some others, but still does better in these tasks. It even beats bigger models on some tests! This shows that training a model just for the task it needs to do (in this case, understanding Indian law) can be more effective than trying to use a big general-purpose language model. |
Keywords
* Artificial intelligence * Bleu * Decoder * Gemini * Language model * Llama * Perplexity * Summarization * Tokenizer * Zero shot
