DuplexWorld: Can voice agents help you get through the day?

Speech-to-speech (S2S) voice agents are increasingly being incorporated into enterprise for customer care and as daily companions for consumers owing to the ease of the conversational modality over text. However, existing benchmarks fail to holistically evaluate voice agents along axes that really matter and are shaped as tests of agentic tool calling against a database. We believe they fail to adequately account for the diversity of conversational dialogue that mundane activities introduce and further, never test how faithfully an agent can assist on tasks that move beyond database manipulation. To tackle this DuplexWorld introduces six worlds where voice agents are especially useful: banking, insurance, travel, healthcare and logistics, and Pathfinding. Agents are evaluated on eleven different types of conversations across 156 scenarios (350+ hours of conversation), each testing conversational and analytical capability to varying degrees. Through extensive evaluation comprising agentic, conversational and speech-naturalness metrics, we show that even the best voice agents leave substantial room for improvement on all 3 axes (Pass@1: 0.490, turn-taking: 0.653, DNSMOS: 3.378). We perform extensive analysis on agentic v conversational performance, world- and conversation type-wise performance, failure modes exploring the explore v exploit lens for Pathfinding conversations and voice agent reliability over all six worlds.

TORUS: A Test of Rendering-Understanding Self-Coherence for Unified Audio Models

Unified audio models capable of audio understanding, audio generation and, increasingly, audio editing are proliferating rapidly. Yet a basic question about them remains unanswered: do the two heads of a unified model agree about the same audio? Current practice evaluates each capability in isolation on specialized benchmarks, and never asks whether a model can make sense of its own generations. We present TORUS, the first self-coherence test for audio-native unified models. TORUS comprises 48 three-stage self-coherence tests carrying 432 six-option questions spanning speech, sound and music across five task families. We holistically evaluate five open unified models alongside a Cascaded Baseline that combines state-of-the-art specialized generation, editing and understanding models. The best unified model answers 50.5% of questions against the Cascaded Baseline's 63.2% and a 16.7% chance floor. Models struggle on audio editing. Among the evaluated audio models (specialized and unified), we observe limited self-coherence, and thus position self-coherence as an essential test for future audio systems.

CaReCoS: A Spectrogram based Visual Benchmark for Cardiac, Respiratory and Cough Sounds

Medical acoustic signals such as respiratory sounds, cardiac auscultations, and cough audio carry rich diagnostic information, yet no existing benchmark evaluates multimodal reasoning over their spectrogram representations. We address both gaps with CaReCoS, a benchmark pairing clinically grounded questions with mel-spectrogram images derived from seven medical audio datasets. Evaluating 9 state-of-the-art vision and omni models, we find that all struggle with fine-grained acoustic features encoded in spectrograms: no model reliably combines visual pattern recognition with medical knowledge, achieving a maximum accuracy of 51.2%, underscoring the need for training on medical sound visualizations.

World Feedback for Clinical Agents: Diagnosing RL in FHIR Environments

Clinical protocol-execution tasks -- checking a lab value, applying a threshold, placing a correctly structured FHIR order -- are natural candidates for RL from world feedback: once clinical SMEs encode decision logic into a verifier, that verifier grades unlimited rollouts without per-episode annotation. But applying RL requires a sound feedback channel and sufficient base capability. We audit MedAgentBench v1/v2, find a 41.7\% silent-finish ceiling that makes inaction the RL dominant strategy, and construct \textbf{MedAgentBench-v3 (MAB-v3)} (508 tasks, 8.9\% ceiling). Training Qwen3-8B exposes two structural barriers: a \emph{capability ceiling} (10/20 task types have 0\% base performance, zero gradient) and a \emph{format-knowledge barrier} (3/20 types require exact clinical codes undiscoverable by exploration). Pure RL reaches 18.2\% pass@1 vs.\ 34.1\% for rule-based SFT; the 15.9~pp gap is attributable entirely to these barriers. A decision/format-knowledge/lookup taxonomy predicts RL learnability and prescribes the fix: SFT to inject codes, RL to learn conditionals.

Beyond Next-Token Prediction: An RLVR Proof of Concept for Tool-Use Agents on Atlassian Workflows

Large language models are trained to predict the next token, not to act inside a specific API. In niche enterprise SaaS workflows -- where success means hitting the right endpoint with the right nested arguments in the right order -- this objective mismatch shows up as silent failures: dropped required fields, hallucinated tools, or early stops after a single read. We ask whether Reinforcement Learning with Verifiable Rewards (RLVR), applied directly in the target environment, closes the gap. As a proof of concept we build a suite of five synthetic environments emulating the Jira REST v3 and Confluence v2 APIs at schema fidelity; rewards are computed entirely from the tool-call trace, with no live API, no learned judge, and no human label in the loop. Scoring prompted Qwen3-1.7B and Qwen3.5-4B on the same checkers that drive GRPO training, we find that on the four scenarios whose rewards are non-degenerate the RL-trained policy lifts average reward from a 4B-baseline range of 0.35--0.92 to 0.95--1.00, with the largest single gain on Confluence page creation ($0.35 \rightarrow 1.00$). We position this as a preliminary step toward outcome-optimised small models for niche enterprise APIs, and foreground two limitations a workshop reader should weigh: hand-crafting verifiable rewards does not scale beyond the handful of endpoints reported here, and one of our five scenarios (ticket-transition) has a saturating reward shape that the prompted 4B already maxes out.

Human + AI for Accelerating Ad Localization Evaluation

Adapting advertisements for multilingual audiences requires more than simple text translation; it demands preservation of visual consistency, spatial alignment, and stylistic integrity across diverse languages and formats. We introduce a structured framework that combines automated components with human oversight to address the complexities of advertisement localization. To the best of our knowledge, this is the first work to integrate scene text detection, inpainting, machine translation (MT), and text reimposition specifically for accelerating ad localization evaluation workflows. Qualitative results across six locales demonstrate that our approach produces semantically accurate and visually coherent localized advertisements, suitable for deployment in real-world workflows.