OmniACBench: A Benchmark for Evaluating Context-Grounded Acoustic Control in Omni-Modal Models
This addresses the need for better evaluation of acoustic control in omni-modal models, which is incremental as it builds on existing multimodal understanding benchmarks.
The authors tackled the problem of evaluating whether omni-modal models can generate appropriate spoken responses by introducing OmniACBench, a benchmark with 3,559 instances covering six acoustic features, and found that eight models performed poorly despite strong textual-output results, with the main bottleneck being multimodal integration for speech generation.
Most testbeds for omni-modal models assess multimodal understanding via textual outputs, leaving it unclear whether these models can properly speak their answers. To study this, we introduce OmniACBench, a benchmark for evaluating context-grounded acoustic control in omni-modal models. Given a spoken instruction, a text script, and an image, a model must read the script aloud with an appropriate tone and manner. OmniACBench comprises 3,559 verified instances covering six acoustic features: speech rate, phonation, pronunciation, emotion, global accent, and timbre. Extensive experiments on eight models reveal their limitations in the proposed setting, despite their strong performance on prior textual-output evaluations. Our analyses show that the main bottleneck lies not in processing individual modalities, but in integrating multimodal context for faithful speech generation. Moreover, we identify three common failure modes-weak direct control, failed implicit inference, and failed multimodal grounding-providing insights for developing models that can verbalize responses effectively.