AI RESEARCH
UT-ACA: Uncertainty-Triggered Adaptive Context Allocation for Long-Context Inference
arXiv CS.LG
•
ArXi:2603.18446v1 Announce Type: cross Long-context inference remains challenging for large language models due to attention dilution and out-of-distribution degradation. Context selection mitigates this limitation by attending to a subset of key-value cache entries, yet most methods allocate a fixed context budget throughout decoding despite highly non-uniform token-level contextual demands. To address this issue, we propose Uncertainty-Triggered Adaptive Context Allocation (UT-ACA), an inference-time framework that dynamically adjusts the context window based on token-wise uncertainty.