conftrace_

autoregression

3 papers

Explore in graph

Co-occurring keywords · papers mentioning both

exponential error accumulation (1) parameter footprint (1) training data requirements (1) eagle (1) latency speedup (1) second-to-top-layer (1) generalist (1) mamba (1) teacher forcing (1) feature uncertainty (1)

Papers

Data-efficient Large Vision Models through Sequential Autoregression ICML 2024 EAGLE: Speculative Sampling Requires Rethinking Feature Uncertainty ICML 2024 The Pitfalls of Next-Token Prediction ICML 2024