arXiv:2608.29613v1 Announce Type: cross Abstract: Function vectors (FVs) have recently emerged as a promising mechanism for steering the behavior of large language models (LLMs) by injecting task-spe...
arXiv:2606. 27717v1 Announce Type: cross Abstract: Prosodic emphasis varies across languages, emotions, and speaking styles, yet existing emphasis detection models are largely trained and evaluated on monolingual neutral read speech.
Emotion recognition in conversation (ERC) is a production capability behind agent-assist prompts, escalation routing, and post-call analytics in contact-center-as-a-service (CCaaS) platforms, where co...
The paper evaluates three approaches for emotion recognition in conversation— a low‑cost stacked ensemble, an off‑the‑shelf LLM prompt, and a confidence‑gated hybrid that escalates only uncertain ensemble predictions to the LLM. Across three datasets (IEMOCAP, MELD, CMU‑MOSI), the hybrid consistently outperforms each pure system, achieving higher weighted F1 scores while routing most traffic through the inexpensive ensemble. This results in significant cost savings (≈$10‑85 per million utterances) and provides an interpretable escalation signal tied to emotion or sentiment shifts.
The paper proposes using multilingual sentiment lexicons to steer language models for sentiment control, especially in low-resource languages. It finds that aggregated multilingual lexicons outperform English-derived steering vectors for mid- and low-resource languages, while monolingual lexicons are still best for languages the model was pre-trained on. The study also explains how representational density, pre-training priors, and architectural biases affect steering effectiveness, noting that large English vectors can cause linguistic collapse in low-resource settings but this is mitigated by norm-dampening in multilingual aggregation.
arXiv:2609.06188v1 Announce Type: new Abstract: Multimodal sentiment analysis and emotion recognition in conversations demand effective modeling of heterogeneous interactions across textual, acoustic...
The paper introduces CASE, a Confusion-Aware Skill Evolution framework designed to improve Multimodal Emotion Recognition (MER) by addressing systematic confusion between semantically adjacent emotions. CASE builds a sparse, self‑evolving library of pair‑specific EmotionSkills that externalize affective expertise from frozen multimodal large language model (MLLM) backbones, allowing continuous refinement and transfer across different models. Experiments on CA‑MER, DFEW, and EMER benchmarks show that CASE significantly boosts accuracy, notably raising Qwen3‑VL‑8B’s performance from 46.80% to 59.13%.
Large language models (LLMs) are increasingly deployed in complex, compound AI systems where performance hinges on the quality of prompts. Recent state-of-the-art optimizers like GEPA (Genetic-Pareto) have argued that reflective instruction evolution can outperform traditional reinforcement learning and few-shot optimization.
The study examines how emotions are represented across layers of large language models (LLMs) by probing eight 1B–9B open‑weight models on three datasets (Twitter, Reddit, autobiographical narratives). It finds that the optimal probing layer varies systematically with the dataset, moving from near‑input layers to deeper layers, and that targeted forward‑pass interventions on these layers degrade performance more than random interventions. Additionally, the selected layers transfer across datasets and emotion categories, and early‑exit representations from these layers outperform full‑depth exits by an average of 6.9 percentage points.
EmoDistill is an offline framework that distills emotional negotiation skills from large language model interactions into smaller agents. It separates emotion selection, handled by an Implicit Q‑Learning selector, from emotion‑conditioned expression, learned by a LoRA‑adapted 7B policy via supervised fine‑tuning and judge policy optimization. Experiments across four negotiation domains show that the full EmoDistill policy outperforms vanilla and IQL‑only baselines, while removing the explicit emotion channel markedly reduces negotiation utility and reveals partial, domain‑dependent transfer to unseen counterparties.
The paper demonstrates that large language model (LLM) evaluators, whether reward‑model based or prompted LLM‑as‑a‑Judge, exhibit significant language bias in multilingual settings. Experiments with semantically identical instruction‑response pairs across 23 languages reveal that lower‑resource languages receive higher scores, a bias that persists across eight open‑weight evaluators and is not detectable by standard pairwise accuracy metrics. The authors link the bias to model uncertainty and language identity, showing it cannot be explained by content difficulty alone.
arXiv:2508.05880v3 Announce Type: replace-cross Abstract: Understanding human emotions is central to user-facing AI applications, safety alignment, and the simulation of human behavior. As emotional...