Tag
AICompanionBench introduces the first publicly available benchmark dataset of 2,123 real-world AI companion conversations annotated across nine safety risk categories, used to evaluate 20 LLMs as safety judges. Results show strong models handle explicit harmful content well but struggle with nuanced risks like manipulation and false positives on benign conversations.
This article presents a structured experiment comparing AI outbound call agents (LuMay Voice Agent, Voxentis, and open-source stacks) for real lead conversion, highlighting their respective strengths in workflow stability, conversational adaptability, and system control.
Introduces DMF, a deterministic memory framework for conversational AI agents that replaces LLM-based compression with classical NLP and mathematical scoring, achieving comparable accuracy to Mem0 while using zero tokens for memory preparation and up to 242× fewer tokens overall.
An analysis of half-duplex vs full-duplex architecture in AI voice models, discussing key features like overlap, backchannels, and barge-in that make voice agents sound robotic.
YouTube launches new conversational search feature "Ask YouTube", allowing users to ask complex questions and refine needs through follow-ups.
Orvera AI, formerly CallBotics, rebranded to reflect enterprise demand shifting from simple voice bots to production-grade AI agent systems that handle workflow execution, governance, and multi-channel orchestration.
Sesame, an AI startup founded by Oculus creators, launched a public preview iOS app with human-like voice agents that can handle natural conversations, parallel searches, and has four distinct AI personalities.
Adobe's Firefly AI Assistant, a conversational agent integrated into design apps like Photoshop and Illustrator, aims to streamline busywork while keeping creative control in users' hands. The beta version shows promising but imperfect results, with a transparent, step-by-step approach to edits.
This paper introduces MeDial-Speech, a dataset of robot-patient and doctor-patient medical dialogues for spoken language processing, and evaluates three LLMs on a sentence selection benchmark, finding Claude Sonnet 4 most accurate.
The article highlights how AI agents are revolutionizing software by eliminating the need to learn tools, and introduces Spoki, an AI conversational platform that manages the entire customer journey across WhatsApp, SMS, and Voice AI, replacing traditional CRM systems.
Octolane is a self-driving AI CRM that you can communicate with conversationally.
A video demonstrates an AI that analyzes user search history to generate an algorithm, enabling it to predict and answer questions before they are asked, showcasing surprisingly predictive capabilities.
Discusses the conversational AI copilot approach for video creation, using Higgsfield supercomputer and Invideo Agent One as examples, and questions whether this orchestrated workflow is more valuable than using underlying models directly.
The article critiques ReAct and Workflow agents for enterprise customer service, highlighting issues like lack of controllability and rigidity, then introduces TeliChat, a code-first conversational agent that separates LLM tasks (intent recognition, NLG) from code-driven business logic for better traceability and debuggability.
This paper proposes a plug-and-play module using self-paced curriculum learning to enhance modality balance in multimodal conversational emotion recognition, achieving consistent F1-score improvements on IEMOCAP and MELD datasets.
The article argues that the most significant recent shift in AI is not about intelligence but memory—AI systems remembering user preferences, habits, and ongoing projects, transforming from mere tools into context-aware assistants.
This paper presents an experimental study investigating whether conversational XAI assistants improve user performance in terms of prediction accuracy, model understanding, and error identification compared to Q&A-based assistance, with preliminary results showing no significant performance differences.
Miso Labs releases Miso TTS 8B, a text-to-speech model based on the Sesame CSM architecture with a Llama 3.2-style backbone, designed for high-quality conversational speech generation and voice continuation.
Google's AI Mode, launched a year ago in the U.S., now has over a billion monthly active users globally and is changing search behavior with longer queries, growing voice/image searches, and increased planning and brainstorming queries.
Google announced Gmail Live, a conversational AI feature powered by Gemini that lets users ask natural language questions about their inbox instead of using traditional search. The feature was introduced at Google I/O and aims to make finding information in emails easier.