3 papers
cs.LG2026
Next Token or Vector? Learning How Far to Decode
Sai Sreenivas Chintha, Chidera Biringa, Rishab Sharma +6
Standard autoregressive decoding advances one token at a time, even when several upcoming tokens can be predicted together. We propose CADENCE (Context-Adaptive DEcoding), a genera…
cs.LG2026
Match the Distribution, Not the Compute: Post-Training Multi-Token Prediction Heads
Prachi Badarayani, Aidan Jay, Chenghui Zhou +8
Multi-token prediction (MTP) improves the throughput of autoregressive generation by enabling the language model to draft multiple next tokens per forward pass, while a verificatio…
cs.LG2026
Can We Trust LLM Judges: A Study of Capability-Dependent Biases and Multi-Judge Ensemble for Bias Calibration
Gemma Zhang, Prachi Badarayani, Asmi Kumar +2
LLMs are increasingly used as automated judges for model training and evaluation, yet individual judges exhibit systematic biases that undermine reliability. Much of prior work has…