HERIT mitigates temporal bias for Hanja-to-English historical translation via RAG pseudo-labeling
DAMI-Lab · hf · 2026-11-06
DAMI-Lab introduces HERIT, adapting machine translation for Hanja historical archives by mitigating temporal bias. The pipeline: (1) a Proposer generates multiple drafts using task instructions and multilingual dictionary examples; (2) retrieval draws historical priors from Hanja-Kor and Hanja-Eng corpora; (3) a Fusor merges drafts with retrieved context into the final sentence; (4) 2-stage fine-tuning combines synthetic and human-labeled data. This improves accuracy across time periods, unlocking global access to remaining Hanja historical texts.
More from Research
- ExPert paper inverts perturbation models to trace cell states back to causal causes — anshulkundaje · 2026-11-06
- Follow-up: gradient flow lets residual states encode information for far-future predictions — mlpowered · 2026-11-06
- Gautam Kamath on AI review: fun science experiment, but false positives can mislead people at scale — thegautamkamath · 2026-11-06
- UCLA's FlowPilot uses human preferences to cut sidewalk robot interventions by 40% — zhoubolei · 2026-11-06
- diffTLM: NeurIPS 2026 Spotlight differentiable vocal tract simulator reconstructs shape from voice — sainingxie · 2026-11-06
- AI model learns thousands of words from one child's audio — only with nightly replay, Science reports — jpillowtime · 2026-11-06