Low-resource languages remain challenging for cross-lingual semantic alignment because of limited parallel corpora. In addition, conventional symmetric alignment may distort the semantic space of a high-resource language through noisy low-resource updates. To address this issue, we propose Monolingual Anchoring for Cross-Lingual Semantic Alignment (MACA), focusing on Uyghur as a low-resource case study. MACA follows an asymmetric paradigm that treats the high-resource language as a fixed semantic anchor and transfers its semantic structure to the Uyghur side. The method consists of three components: (1) Anchored Embedding Initialization for newly introduced Uyghur subwords, (2) Cross-Lingual Neighborhood Anchoring for structural alignment between Uyghur and the anchor language, and (3) Monolingual Structure Anchoring for improving the internal semantic organization of Uyghur representations. Experiments centered on Uyghur-Chinese show that MACA outperforms LaBSE, the strongest off-the-shelf multilingual baseline in our comparison, by 7.67 points on cross-lingual STS. In an exploratory Uyghur-English zero-shot setting, MACA also surpasses LaBSE by 2.45 points without using Uyghur-English training data. These results provide evidence for the effectiveness of MACA in the evaluated Uyghur setting and suggest that monolingual anchoring may be further explored for related low-resource languages, such as Kazakh, Kyrgyz, and Uzbek.
Ruohao Yan, Huaping Zhang, Yuwen Niu et al.· Journal of King Saud Univers...· 0 citations
Large language models (LLMs) increasingly serve as data-driven reasoners, yet their chains-of-thought (CoT) can be unfaithful even when final answers are correct. Most existing ''verification'' signals are not diagnostic: answer matching observes only the outcome, LLM-as-judge provides subjective and non-verifiable critiques, and scalar rewards (e.g., PRMs/RMs) offer little insight into where a multi-step derivation fails.We propose SymDiag, a neuro-symbolic framework that reframes reasoning verification as structured failure diagnosis. SymDiag translates natural-language CoT into symbolic constraints and performs step-level satisfiability/entailment checks to (i) localize failing steps and (ii) produce verifiable diagnostic evidence, including counterexamples, inconsistency witnesses, and missing-premise indicators. A central challenge is that apparent ''logic violations'' can be caused either by genuine reasoning defects or by neural-to-symbolic translation noise. SymDiag therefore incorporates a Self-Auditor that disentangles TranslationError from ReasoningError via dual symbolic encodings consistency checks, enabling robust diagnosis under partial observability. Across diverse mathematical, logical, scientific, and general reasoning benchmarks, SymDiag improves detection of unfaithful reasoning and provides substantially more effective feedback for multi-round reasoning repair than outcome-only verification and LLM-based judging, offering a principled foundation for trustworthy and scalable reasoning diagnosis.
Wenyao Cui, Huaping Zhang, Yongyi Huang et al.· Proceedings of the 32nd ACM...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
