WattBot 2026: Investigating AI’s Environmental Footprint with RAG

WattBot is back for the 2026 Machine Learning Marathon (MLM26) – bigger and tougher than last year. Teams build retrieval-augmented generation (RAG) systems that answer over 500 questions about AI’s environmental impact from a corpus of over 100 papers and public reports, including arXiv preprints, peer-reviewed work, LBNL/IEA/GAO reporting, corporate sustainability filings, and a regional cluster on Wisconsin and the Great Lakes. Every answer needs a citation and supporting evidence, and when the corpus can’t support an answer, the system must say so instead of guessing. The goal: turn scattered evidence into transparent, actionable answers for researchers, engineers, and policymakers.

Challenge design

• Task: For each question, return a concise answer, a normalized answer value, the supporting document ID(s), verbatim supporting materials, and the reasoning connecting evidence to answer – or explicitly abstain (is_blank) when the corpus can’t answer.
• Question types: Range from locating a stated fact, to reading values off figures (OCR and vision models are fair game), combining documents, attributing a company’s claim as a claim, and reconciling conflicting sources.
• Evaluation: The WattBot Score – a weighted accuracy over answer correctness (0.75), citation F1 (0.20), and proper abstention on unanswerable questions (0.05). The exact scorer ships with the data so you can validate locally before submitting.
• Bonus track: Teams are invited to build a chatbot application around their RAG methods – judged informally on visible citations, speed, grounded accuracy, and completeness.

Links

• Kaggle challenge: WattBot 2026

Related resources

• Project: WattBot 2025: Estimating AI Emissions with RAG: The first WattBot challenge, including the winning RAG solution and deployment repos – a natural starting point for 2026 teams.
• Talk: Deploying RAG in Bedrock vs. Local: WattBot 2025 Case Study: ML+X forum where the winning WattBot 2025 RAG system was deployed in AWS Bedrock and locally with open-source models.
• Talk: AI’s Environmental Footprint: Insights and Actions: The ML+X forum where WattBot was first introduced alongside LLM energy benchmarking work.
• Notebook: Exploring Fact-Based QA with RAG: Romeo and Juliet: Build an end-to-end RAG pipeline from scratch – a great starting point before tackling WattBot.
• Workshop: Intro to AWS for ML/AI: Covers AWS SageMaker and Bedrock for cloud-based ML/AI workflows, including RAG deployment.
• Notebook: Understanding Quantization and Precision: Learn how quantization (e.g., 4-bit) reduces model size and memory requirements – useful for running open models locally in your pipeline.
• Paper: KohakuRAG: A simple RAG framework with hierarchical document indexing: The WattBot 2025 winning team’s writeup – hierarchical document indexing, an LLM-powered query planner, and abstention-aware ensemble voting, with ablations of what drove the score. Released as open-source software and required reading for 2026 teams.

Comments