AI Paper Atlas

Research map / Label-efficient and robust learning / Knowledge distillation

CTPD: Cross Tokenizer Preference Distillation

Truong Nguyen, Phi Van Dat, Ngan Nguyen, Linh Ngo Van, Trung Le, Thanh Hong Nguyen

AAAI 2026 · Label-efficient and robust learning › Knowledge distillation · approach: Rethinking distillation objectives · 1 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
use CTPD instead of TIS-DPOLlama-3.1-8B student · Six-benchmark averageaccuracy (%): 66.16 -> 67.42Improved
use CTPD instead of TIS-DPOLlama-3.2-1B student · Six-benchmark averageaccuracy (%): 42.60 -> 43.26Improved
use CTPD instead of TIS-DPOLlama-3.1-8B student · GSM8kaccuracy (%): 54.31 -> 57.47Improved
use CTPD instead of TIS-DPOLlama-3.1-8B student · TruthfulQAaccuracy (%): 53.86 -> 55.22Improved
use CTPD instead of cross-tokenizer KD (Multi-Level OT)Llama-3.1-8B student · Six-benchmark averageaccuracy (%): 64.58 -> 67.42Improved
use CTPD instead of DSKDLlama-3.2-1B student · Six-benchmark averageaccuracy (%): 41.12 -> 43.26Improved
add teacher-estimated span weights instead of random weights (ablation)CTPD · Six-benchmark averageaccuracy (%): 54.80 -> 67.42Improved
teacher-estimated weights instead of length-averaged weights (ablation)CTPD · Six-benchmark averageaccuracy (%): 65.47 -> 67.42Improved
teacher-estimated weights instead of student-estimated weights (ablation)CTPD · Six-benchmark averageaccuracy (%): 65.88 -> 67.42Improved
teacher-estimated weights instead of teacher-student SFT estimate (ablation)CTPD · Six-benchmark averageaccuracy (%): 64.51 -> 67.42Improved
use teacher instead of student as reference model (ablation)CTPD · Six-benchmark averageaccuracy (%): 65.27 -> 67.42Improved
use CTPD instead of DPOLlama-3.1-8B student · MMLUaccuracy (%): 65.26 -> 66.65Mixed

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