Research map / Vision-language and multimodal / Multimodal LLMs
Explainable Oracle Bone Script Recognition via Multimodal Pictographic Reasoning
Yin Wu, Zhengxuan Zhang, Jiayu Chen, Chang Xu, Yuyu Luo, Nan Tang, Hui Xiong
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes an explainable oracle bone script recognition approach using multimodal pictographic reasoning, moving beyond opaque classification of glyphs.
Problem
Oracle bone script recognition systems are accurate classifiers but treat glyphs as opaque data, ignoring their pictographic nature and failing to make recognition explainable to the public.
Approach
Frames oracle bone script recognition as multimodal pictographic reasoning so that predictions come with explanations grounded in the glyphs' visual metaphors (abstract only; method details not available).
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add stage 2 multimodal adaptation (glyph and real-world images) after domain adaptation (ablation) | OBS-VM · OracleVis test set | Pictographic Rationality (human): 2.31 -> 3.79 | Improved |
| add stage 2 multimodal adaptation (ablation) | OBS-VM · OracleVis test set | Top-1 accuracy: 0.13 -> 0.23 | Improved |
| add stage 2 multimodal adaptation (ablation) | OBS-VM · OracleVis test set | Top-5 accuracy: 0.16 -> 0.33 | Improved |
| two-stage fine-tuning on OracleVis | Qwen2-VL-7B · OracleVis test set | Pictographic Rationality (human): 2.37 -> 3.79 | Improved |
| fine-tune Qwen2-VL-7B on OracleVis instead of using GPT-4o | OBS-VM · OracleVis test set | Top-1 accuracy: 0.17 -> 0.23 | Improved |
| fine-tune on OracleVis vs GPT-4o | OBS-VM · OracleVis test set | Semantic Rationality (human): 4.29 -> 4.33 | Improved |
| OBS-VM vs human expert-style explanations | OBS-VM · 50 shared glyphs | Pictographic Rationality: 4.20 -> 3.78 | Hurt |
| OBS-VM vs human explanations on readability | OBS-VM · 50 shared glyphs | Readability: 4.30 -> 4.28 | No clear effect |
| interactive learning with OBS-VM vs static web resources | non-expert learners · User study (N=24) | Understanding gain (/10): +1.7 -> +5.5 | Improved |
| interactive learning with OBS-VM vs static web resources | non-expert learners · User study (N=24) | Interest gain (/5): +0.4 -> +1.9 | Improved |
| interactive learning with OBS-VM vs static web resources | non-expert learners · User study (N=24) | Confidence gain (/5): +0.3 -> +2.0 | Improved |
Closest papers by meaning
- OBI-Bench: Can LMMs Aid in Study of Ancient Script on Oracle Bones? (ICLR 2025 · similarity 0.92)
- Making Visual Sense of Oracle Bones for You and Me (CVPR 2024 · similarity 0.89)
- Component-Level Segmentation for Oracle Bone Inscription Decipherment (AAAI 2025 · similarity 0.88)
- VaseVQA-3D: Benchmarking 3D VLMs on Ancient Greek Pottery (ICLR 2026 · similarity 0.86)
- ProtoSnap: Prototype Alignment For Cuneiform Signs (ICLR 2025 · similarity 0.85)
- EpiAgent: An Agent-Centric System for Ancient Inscription Restoration (CVPR 2026 · similarity 0.84)
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.
