The Flow has not summarised this story yet — read it at arXiv Computation and Language.
The study investigates how multimodal large language models (MLLMs) use prosodic cues in sarcasm detection. By testing Qwen2.5‑Omni and Qwen3‑Omni on Mandarin Chinese and English across five modality conditions, the authors find that adding audio increases false positives without improving true positives. Acoustic error analysis shows that models rely on a stereotypical prosodic pattern—elevated pitch and irregular pausing—that does not align with genuine sarcasm cues, and manipulating these dimensions alone can raise false positive rates up to 60%. The same effect appears in Gemini 3 Flash Preview, indicating the heuristic is not limited to a single architecture.
Multimodal Large Language Models (MLLMs) process speech and text jointly, yet whether they exploit prosodic cues for pragmatic inference or rely on surface acoustic patterns has received little system...
arXiv:2606. 06168v1 Announce Type: new Abstract: We present ProSarc, an audio-only framework that detects sarcasm by modelling temporal prosodic incongruity, that is, the mismatch between local prosodic dynamics and the utterance-level emotional baseline.
DEAF (Diagnostic Evaluation of Acoustic Faithfulness) is a benchmark comprising over 2,700 conflict stimuli that test Audio Multimodal Large Language Models (Audio MLLMs) across three acoustic dimensions: emotional prosody, background sounds, and speaker identity. The benchmark uses a controlled multi-level evaluation framework that gradually increases textual influence, from semantic conflicts to misleading prompts, to separate content-driven bias from prompt-induced sycophancy. Diagnostic metrics are introduced to measure how much models rely on textual cues versus acoustic signals, and an evaluation of seven Audio MLLMs shows that while models respond to acoustic variations, their predictions are largely driven by textual inputs, highlighting a gap between benchmark performance and true acoustic understanding.
SpeechCritic learns a diagnostic speech judge from only about 300 human-labeled comparisons, calibrating a frontier audio-language model with non-binding hints that select acoustic measurements agreeing with human judgments. The approach improves dimension-level agreement by 6.3 points and reduces mismatch in Tie rates by 10.4 points, and training a 7B judge with SFT, OPD, and RL yields distinct behaviors, with RL producing more specific rationales. The method works across language pairs, demonstrating a path from limited human preferences to a diagnostic speech judge.
Prosodic cues can convey task-relevant information that alters the trajectory and outcome of a task-oriented dialogue, even when the words themselves remain unchanged. Yet existing benchmarks typically evaluate prosodic perception, response appropriateness, and task-oriented dialogue in isolation, making it difficult to test whether prosodic evidence changes downstream decisions.