The paper investigates how dense embedding models can be used for stance-aware argument retrieval, a task that requires both topic relevance and correct stance (support or attack) toward a claim. Experiments reveal that current models favor topical overlap and ignore stance, and that contrastive training to fix this bias leads to over-correction, where models focus too much on polarity keywords at the expense of topic relevance. To address this, the authors propose diagnostic word-ablation metrics and a data‑centric solution involving a balanced argument curriculum and LLM‑augmented stance‑inverted arguments, which helps powerful models learn deeper directional logic and improves stance‑aware retrieval performance.
By Angelo Sparacino, Francesca Toni, Adam Dejl
The paper evaluates Jev, a specialized decision model, on stance detection tasks using the VAST (English) and ZS-CSD (Chinese) datasets. Jev matches GPT‑5.6 and outperforms other general‑purpose LLMs on VAST, but lags behind stronger LLMs on ZS‑CSD, especially in distinguishing favor from against. The authors suggest that Jev’s limitations may stem from difficulties in understanding reply relationships and stance direction rather than conversation length.
By Xing Li, Jinzhong Ning, Yijia Zhang, Liang Yang, Hongfei Lin
arXiv:2603. 23841v2 Announce Type: replace-cross Abstract: While Large Language Models (LLMs) are increasingly used as primary sources of information, their potential for political bias may impact their objectivity.
By Rohan Khetan, Ashna Khetan
arXiv:2607. 24191v1 Announce Type: cross Abstract: Conversational stance detection has shifted from static text analysis to dynamic multimodal modeling.
By Heyan Chai, Xin Li, Wenjie Wang, Jianyang Qin, Chaoyang Li, Lu Wang, Hao Chen, Qing Liao
MMDS-Bench is a new diagnostic benchmark for multimodal dynamic stance classification in social media parent‑reply interactions. It contains 3,482 multimodal instances annotated with a seven‑label stance taxonomy, plus an 800‑instance subset that demands structured reasoning over parent and reply understanding and stance‑relation inference. The benchmark also tags each instance with five challenge factors—multimodal fusion, parent framing, non‑literal expression, interaction reasoning, and label‑boundary ambiguity—and evaluates 12 multimodal large language models using a reference‑grounded LLM‑judge protocol, revealing that current models still struggle with relational inference beyond separate parent and reply comprehension.
By Yuzhe Ding, Kang He, Li Zheng, Shengwu Zheng, Teng Shi, Fei Li, Chong Teng, Donghong Ji
arXiv:2606.06443v3 Announce Type: replace
Abstract: Large language models are increasingly used to simulate social media users and infer how individuals may respond to online discussions. However, it...
By Xinnong Zhang, Wanting Shan, Hanjia Lyu, Zhongyu Wei, Jiebo Luo