New AI model TellTale recognizes ambivalence from video transcripts

Researchers have developed TellTale, a novel approach for recognizing ambivalence and hesitancy in video interviews using only text transcripts. The system combines three probability streams: two text encoders fine-tuned with LoRA adapters and a zero-shot instruction LLM. TellTale achieved a Macro-F1 score of 0.7364 on a private test set, significantly outperforming the official vision-based baseline. AI

IMPACT This research demonstrates a novel method for analyzing human emotion and intent from text, potentially improving automated content moderation and user behavior analysis.

RANK_REASON The cluster contains an academic paper detailing a new AI model and its performance on a specific task. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• 11th ABAW Workshop
• 3rd A/H Video Recognition Challenge
• Abdel-Karim Al-Tamimi
• BAH dataset
• ECCV 2026
• LoRA
• mDeBERTa-v3-base
• multilingual-e5-large

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →