.png)
KiranaChain is a large scale behavioral intelligence benchmark designed to model decentralized retail negotiations within the Indian Kirana ecosystem. The dataset contains 1000000 turn level records grouped into 125000 complete negotiation trajectories spanning distributor retailer interactions across 12 Indian states. The dataset captures informal credit allocation mechanisms Udhar dynamics trust score evolution liquidity stress indicators supply chain disruptions environmental externalities multilingual bargaining behavior and structured negotiation telemetry. Each trajectory consists of 8 sequential negotiation turns enabling trajectory based research in Reinforcement Learning Multi Agent Systems Behavioral Economics Credit Risk Modeling Graph Learning and Conversational AI. KiranaChain contains 40 structured features organized across agent profiles financial indicators environmental conditions and negotiation state variables. The generation framework combines Gaussian Copula based dependency modeling Non Stationary Markov negotiation processes and Gaussian Process behavioral perturbations to create statistically consistent large scale synthetic negotiation data. The dataset is released by Atlas AI Labs and developed by undergraduate researchers (Dhadi Sai Praneeth Reddy, Mididuddi Dhatri, Biradar Amulya) from Vasavi College of Engineering Autonomous Hyderabad India. See the full description on the dataset page: https://huggingface.co/datasets/Atlas-AI-Labs/KiranaChain.
© 2026 - Copyright AIKosh. All rights reserved.