Our paper, CG-TTRL: Context-Guided Test-Time Reinforcement Learning for On-Device Large Language Models, co-authored by QMUL PhD student Peyman Hosseini and Networks Research Group member Ignacio Castro, has been published in TACL and selected for presentation at EMNLP 2026. The work proposes a context-guided approach to improve the accuracy and efficiency of test-time reinforcement learning for on-device LLMs.
Reference: Peyman Hosseini, Ondrej Bohdal, Taha Ceritli, Ignacio Castro, Matthew Purver, Mete Ozay and Umberto Michieli. CG-TTRL: Context-Guided Test-Time Reinforcement Learning for On-Device Large Language Models. Transactions of the Association for Computational Linguistics (TACL), 2026.
Paper on on-device LLMs published in TACL