Adapting llm to multi-lingual esg impact and length prediction using in-context learning and fine-tuning with rationale

Ashvini Kumar Jindal, Pawan Kumar Rajpoot, Ankur Parikh

Abstract: The prediction of Environmental, Social, and Governance (ESG) impact and duration (length) of impact from company events, as reported in news articles, hold immense significance for investors, policymakers, and various stakeholders. In this paper, we describe solutions from our team "Upaya" to ESG impact and length prediction tasks on one such dataset ML-ESG-3. We employed two different paradigms to adapt Large Language Models (LLMs) to predict both ESG impact level and length of events. In the first approach, we leverage GPT-4 within the In-context learning (ICL) framework where a retriever identifies top K-relevant in-context learning examples for a given test example. The second approach involves instruction-tuning Mistral (7B) LLM to predict impact level and duration, supplemented with rationale generated using GPT-4. Our models secured second place in both French tasks where for one task fine-tuned Mistral model outperformed and for other task, GPT-4 with ICL outperformed. These results demonstrate the potential of different LLM-based paradigms for delivering valuable insights within the ESG investing landscape.