AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Token-Scaled Logit Distillation for Ternary Weight Generative Language Models

Minsoo Kim, Sihwa Lee, Janghwan Lee, Sukjin Hong, Du-Seong Chang, Wonyong Sung, Jungwook Choi

NeurIPS 2023 · Large language models › Language, safety and interpretability · approach: LLM unlearning and benchmarks · 2 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace Logit KD with token-scaled logit distillation (TSLD)W2A16 QAT of OPT-6.7B · PTBperplexity: 11.33 -> 11.17Improved
add ground-truth loss naively to Logit KD (Logit+GT) (ablation)W2A16 QAT of OPT-1.3B · PTBperplexity: 14.80 -> 16.41Hurt
add ground-truth loss naively to Logit KD (Logit+GT) (ablation)W4A16 QAT of OPT-1.3B · PTBperplexity: 13.73 -> 15.03Hurt
scale Logit+GT with per-token teacher cross-entropy weights (TSLD) (ablation)W2A16 QAT of OPT-1.3B · PTBperplexity: 16.41 -> 14.60Improved
add layer-to-layer (L2L) distillation to Logit KD (ablation)W2A16 QAT of OPT-1.3B · PTBperplexity: 14.80 -> 17.62Hurt
TSLD instead of Logit KDOPT-6.7B channel-wise W2 · GSM8Kaccuracy: 21.08 -> 24.49Improved
TSLD instead of Logit KDLLaMA-7B channel-wise W2 · GSM8Kaccuracy: 25.47 -> 26.23Improved
TSLD instead of Logit KDOPT-2.7B W2A16 · ARC_challengeaccuracy: 31.91 -> 33.45Improved
TSLD instead of Logit KDOPT-2.7B W2A16 · GSM8Kaccuracy: 20.02 -> 20.24No clear effect
TSLD instead of Logit KDOPT-1.3B W2A16 · GLUE CoLAaccuracy: 48.72 -> 54.47Improved
QAT with TSLD instead of 4-bit PTQ (OPTQ)OPT-1.3B W4A16 · GLUE RTEaccuracy: 56.32 -> 75.97Improved
channel-wise instead of tensor-wise quantization (ablation)TSLD W2A16 on GPT-Neo-1.3B · PTBperplexity: 30.60 -> 19.27Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy