AI Paper Atlas

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

Teaching Metric Distance to Discrete Autoregressive Language Models

Jiwan Chung, Saejin Kim, Yongrae Jo, Jaewoo Park, Dongjun Min, Youngjae Yu

ICLR 2026 · Large language models › Language, safety and interpretability · approach: Sign and cross-lingual translation · 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
add place value weighting (ablation)DIST2Loss (Llama-3.2-1B) · meta linear regressionMAE: 0.098 -> 0.092Improved
add contrastive loss (ablation)DIST2Loss (Llama-3.2-1B) · meta linear regressionMAE: 0.099 -> 0.092Improved
add distance-aware target (vs label smoothing 0.1) (ablation)DIST2Loss (Llama-3.2-1B) · meta linear regressionMAE: 0.099 -> 0.092Improved
use a random distance metric instead of Euclidean (ablation)DIST2Loss · reward modelingaccuracy (%): 85.3 -> 76.0Hurt
add DIST2Loss to SFTLlama-3.1-8B · RewardBench averageaccuracy (%): 75.3 -> 85.3Improved
add DIST2Loss to SFTPhi-3.5-vision · RefCOCO+ testBaccuracy (%): 78.7 -> 81.4Improved
add vocabulary-based soft target instead of DIST2LossPhi-3.5-vision · RefCOCO+ testAaccuracy (%): 91.6 -> 90.6Hurt
add DIST2Loss to SFTLLaRA · VIMABench L2 (1K data)accuracy (%): 46.2 -> 51.5Improved
add DIST2Loss to SFTLlamaGen 111M · ImageNet 256 (50 epochs)FID: 10.03 -> 9.41Improved
add DIST2Loss to SFTLlamaGen 343M · ImageNet 256 (300 epochs)FID: 3.08 -> 3.04Improved
set loss weight alpha to 0.1 (vs 1.0) (ablation)DIST2Loss · reward modelingaccuracy (%): 77.3 -> 85.3Improved
set loss weight alpha to 0.001 (vs 0.1) (ablation)DIST2Loss · reward modelingaccuracy (%): 85.3 -> 75.6Hurt

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