This paper investigates how the final prompt in conversational AI-search evaluations differs from the conversation history, using two corpora of commercial and PRISM conversations.

The paper introduces TurnGate, a response-aware defense mechanism that detects the earliest turn in a multi-turn dialogue where the accumulated interaction enables a harmful action, significantly impr…

This paper revisits the assumption that address in multi-party dialogues is discrete and explores it as a continuous phenomenon using a human dialogue corpus.

This paper proposes a multi-turn retrieval-augmented generation pipeline for conversational systems across four domains.

This paper explores the effect of conversational timing properties on automatic speech recognition (ASR) systems by controlling and optimizing pause and overlap timing distributions.

This paper investigates entrainment effects in multi-party human-agent conversations and finds limited global entrainment and cohort-dependent entrainment with the agent.

The paper introduces PersTurnBench, a novel benchmark and evaluator for assessing personalized user conversation satisfaction at specific turns, addressing the limitation of generic response quality m…

The paper proposes scheduling LLM agent workloads at the conversation level rather than the turn level, significantly reducing latency and improving energy efficiency by transforming unpredictable mul…

This paper introduces model merging as a training-free strategy for designing a single retrieval model that operates across both ad-hoc and conversational settings, improving ad-hoc search capabilitie…

The paper introduces Dialogs, a new Russian conversational speech corpus with high-quality recordings, segmented utterances, and expressive prosody labels.

The paper introduces VIP-MINGLE, a multimodal dataset of 59 hours of group conversations in both in-person and videoconferencing settings, with raw data, psychometric data, processed features, and ann…

This paper introduces BayLing-Duplex, a native full-duplex Speech Language Model that decides when to listen, speak, and stop without relying on an external Voice Activity Detection module.

CA-BED is a novel framework that improves LLM performance in interactive question-answering by integrating Bayesian Experimental Design to strategically select questions that maximize information gain…

The paper introduces RHELM, a new benchmark designed to test LLMs' long-term memory by simulating realistic, complex, and evolving dialogues that integrate multiple heterogeneous data sources.

THRD introduces a novel, training-free framework that models temporal risk accumulation to effectively defend against multi-turn jailbreak attacks on LLMs, significantly reducing attack success rates…

The paper introduces TikStance, a multimodal and context-aware dataset for stance detection in political discussions on TikTok.

This paper introduces Latent-IM, a framework for controlling conversational moves in large language models for dialogue systems.