Multiple Choice

In a speculative decoding process, the number of consecutively accepted tokens from the start of a draft sequence, denoted by nan_a, is determined by finding the index of the first rejected token. The formula is: na=min{t11tτ,rt>p(y^i+t)q(y^i+t)}n_a = \min \{t-1 | 1 \le t \le \tau, r_t > \frac{p(\hat{y}_{i+t})}{q(\hat{y}_{i+t})} \}.

Given a draft sequence of 5 tokens (τ=5\tau=5) with the following randomly generated numbers (rtr_t) and probability ratios (p/qp/q), what is the calculated value of nan_a?

  • t=1: r1r_1=0.4, p/qp/q=0.7
  • t=2: r2r_2=0.8, p/qp/q=0.9
  • t=3: r3r_3=0.6, p/qp/q=0.5
  • t=4: r4r_4=0.3, p/qp/q=0.8
  • t=5: r5r_5=0.7, p/qp/q=0.6

0

1

Updated 2025-10-03

Contributors are:

Who are from:

Tags

Ch.5 Inference - Foundations of Large Language Models

Foundations of Large Language Models

Foundations of Large Language Models Course

Computing Sciences

Analysis in Bloom's Taxonomy

Cognitive Psychology

Psychology

Social Science

Empirical Science

Science