yagao403/llama3.1-70B-memento-no-more-OfficeBench

Llama-3.1-70B-Instruct + OfficeBench (Finetuned)

This model is based on Llama-3.1-70B-Instruct, fine-tuned on the OfficeBench for multi-step tool-use office tasks.

Training Details

• —Dataset: OfficeBench – an office automation benchmarks for evaluating current LLM agents' capability to address office tasks in realistic office workflows.
• —Training Framework: Memento-No-More – a novel framework for teaching models to internalize hints and perform multi-skill reasoning.
• —Fine-tuning Rounds: 3
• —Model Base: Llama-3.1-70B-Instruct

Reference

For detailed information on the training methodology, architecture, and evaluations, please refer to our paper:

Alakuijala, M., Gao, Y., Ananov, G., Kaski, S., Marttinen, P., Ilin, A., & Valpola, H. (2025). Memento No More: Coaching AI Agents to Master Multiple Tasks via Hints Internalization. arXiv preprint arXiv:2502.01562.