Read our full paper here: https://arxiv.org/abs/2511.10684 Authors: Anupama Sitaraman, Bharathan Balaji, Yuvraj Agarwal
** Abstract ** Investigating the effects of climate change and global warm- ing caused by GHG emissions have been a key concern worldwide. These emissions are largely contributed to by the production, use and disposal of consumer products. Thus, it is important to build tools to estimate the environmental im- pact of consumer goods, an essential part of which is con- ducting Life Cycle Assessments (LCAs). LCAs specify and account for the appropriate processes involved with the pro- duction, use, and disposal of the products. We present Spider- Gen, an LLM-based workflow which integrates the taxonomy and methodology of traditional LCA with the reasoning ca- pabilities and world knowledge of LLMs to generate graphi- cal representations of the key procedural information used for LCA, known as Product Category Rules Process Flow Graphs (PCR PFGs). We additionally evaluate the output of Spider- Gen by comparing it with 65 real-world LCA documents. We find that SpiderGen provides accurate LCA process informa- tion that is either fully correct or has minor errors, achieving an F1-Score of 65% across 10 sample data points, as com- pared to 53% using a one-shot prompting method. We ob- serve that the remaining errors occur primarily due to differ- ences in detail between LCA documents, as well as differ- ences in the “scope” of which auxiliary processes must also be included. We also demonstrate that SpiderGen performs better than several baselines techniques, such as chain-of- thought prompting and one-shot prompting. Finally, we high- light SpiderGen’s potential to reduce the human effort and costs for estimating carbon impact, as it is able to produce LCA process information for less than $1 USD in under 10 minutes as compared to the status quo LCA, which can cost over $25000 USD and take up to 21-person days.
This source code contains the following:
The spiderGen folder contains the core modules to run SpiderGen, including:
- utils, which contains code to configure an LLM backend for SpiderGen
- procedure_generation, which contains the prompts for each stage of the workflow, as well as code for the clustering steps. The proceedure generation steps are based on ISO standard ISO 14025:2006.
- spidergen_example.ipynb, which is Jupyter notebook containing an example PFG using the SpiderGen workflow
To run, ensure that you have all of the requirements in your environment and edit config.py with your api key and model configuration. You can select between openai, anthropic and ollama models.
To ensure that requirements are installed run "pip install -r requirements.txt"
Then, run each cell to get the SpiderGen output!
Below: Coming soon!! The evaluation folder contains the modules used to evaluate SpiderGen, including:
- pmi.py, which contains code for calculating the PMI between ground-truth PCR PFG processes and SpiderGen PFG processes
- metrics.py which contains code for calculating existing semantic similarity metrics, such as ROUGE and BLEU between ground-truth PCR PFG processes and SpiderGen PFG processes
- baselines.py which contains LLM prompts and code to generate the baselines used in our paper
To get started with this codebase, first ensure that all of the requirements are installed. These requirements are listed in requirements.txt. To install these requirements, create a new python environment and run