Deciphering scanned Oracle Bone Script (OBS) images into corresponding modern Chinese characters is a crucial step in AI-assisted OBS decipherment. As a highly pictographic writing system, OBS often embeds semantic information in glyph forms, making its decipherment depend on more than visual similarity alone. However, how to effectively exploit such glyph-meaning cues and how to make the decipherment process interpretable remain challenging. To this end, we propose an LVLM-based framework that bridges glyphs and meanings through structured glyph analysis. Specifically, we introduce a progressive training strategy that guides the model from radical analysis to pictographic analysis and finally to mutual analysis, enabling it to capture semantic information embedded in OBS glyphs. The generated textual analyses serve as both interpretable reasoning traces and explicit evidence for retrieving corresponding modern Chinese characters from a dictionary. We also construct the Pictographic Decipherment OBS Dataset (PD-OBS), which provides detailed radical and pictographic analyses to support glyph-analysis learning and dictionary-based retrieval. Experiments on OBS benchmarks demonstrate the strong zero-shot capability of our method, especially its substantially improved Top-10 accuracy on unseen OBS, highlighting its ability to provide reliable and interpretable candidate references for AI-assisted archaeological and historical studies. The code and dataset will be released at https://github.com/PKXX1943/PD-OBS.

Chinese ancient documents, invaluable carriers of millennia of Chinese history and culture, hold rich knowledge across diverse fields but face challenges in digitization and understanding—traditional methods only scan images, while current Vision-Language Models (VLMs) struggle with their visual/linguistic complexity. Existing document benchmarks focus on English printed texts or simplified Chinese, leaving a gap for evaluating VLMs on ancient Chinese documents. To address this, we present AncientDoc, the first benchmark for Chinese ancient documents, designed to assess VLMs from OCR to knowledge reasoning. AncientDoc includes five tasks (page-level OCR, vernacular translation, reasoning-based QA, knowledge-based QA, linguistic variant QA) and covers 14 document types, over 100 books, and about 3,000 pages. Based on AncientDoc, we evaluate mainstream VLMs using multiple metrics, supplemented by a human-aligned large language model for scoring.

Recent work shows that context can both amplify and mitigate social bias in LLMs. Yet a key question remains underexplored: when context provides counter-stereotypic evidence, is that evidence taken up equitably across demographic targets? Using agency–communality bias as a testbed, we introduce a sequential evaluation framework that holds total counter-evidence fixed while varying its evidence structure, adapting the bookkeeping–subtyping contrast from social cognition. Across three instruction-tuned LLMs evaluated in English, counter-evidence is taken up in explicit group-level judgement and transfers to explicit judgement about held-out individuals. The prior-category structure of this uptake, however, does not carry over to generation: communal-prior targets lose their correction advantage, absorb evidence more slowly, and the weakest uptake repeatedly appears for Asian and Hispanic female targets within our controlled target set. The tested prompt instruction and DPO fine-tuning variants yield partial improvements at prior-category level but do not reliably repair these target-level gaps. These findings show that fairness audits for context-rich LLM systems should go beyond checking whether corrective evidence is present and test whether models use that evidence equitably in downstream generation.