Multimodal large language models (MLLMs) perform well across many tasks but often exhibit Western bias, limiting their effectiveness in regions such as Southeast Asia (SEA). In this work, we study how both evaluation and model design can be better adapted to the linguistic and cultural characteristics of this region. We present the SEA-Omni Benchmark Suite, which brings together a set of SEA-focused image and video datasets capturing phenomena such as intra-sentential code-switching and culturally grounded multimodal cues. Alongside these benchmarks, we develop MERaLiON-Omni-10B, an omni-perception model trained with a progressive, data-efficient pipeline that moves from vision-language alignment to more advanced multimodal fusion. To better account for region-specific cultural signals, we study text-mediated cultural alignment in SEA multimodal settings using GRPO to surface culturally distinctive responses that are often suppressed by generic decoding strategies. We introduce a fact-anchored dual-reward mechanism to balance cultural richness with multimodal consistency in which cultural density is measured via localized knowledge entities of generated texts, and multimodal coherence is assessed against silver textual descriptions of the multimodal input. Experimental results indicate that the resulting model performs strongly on SEA benchmarks while maintaining reasonable general capabilities.

The rapid advancement of audio and multimodal large language models has unlocked transformative speech understanding capabilities, yet evaluation frameworks remain predominantly English-centric, leaving Southeast Asian (SEA) languages critically underrepresented. We introduce SEA-SpeechBench, to the best of our knowledge, the first large-scale multitask benchmark that evaluates speech understanding in 11 SEA languages through 97,194 samples across 99 evaluation sets and 597 hours of curated audio data. Our benchmark comprises 9 diverse tasks across 3 categories: speech processing (automatic speech recognition, speech translation, spoken question answering), paralinguistic analysis (emotion, gender, age, speaker recognition), and temporal understanding, a novel dimension featuring timestamped content queries and temporal localization within extended audio sequences up to 3 minutes. We implement multilingual prompting in both native SEA languages and English to reflect user interactions with audio-language models. Evaluation of leading open-source and proprietary systems reveals marked performance gaps. Across all models, performance remains underwhelming on temporal understanding, emotion recognition, and speech translation. Prompting in low-resource languages such as Burmese and Tamil lags behind English by up to 41 percentage points. Our findings expose critical model limitations and underscore the need for inclusive model development. The SEA-SpeechBench benchmark is available at https://zwenyu.github.io/SEA-SpeechBench.

Recent event extraction (EE) methods rely on pre-trained language models (PLMs) but still suffer from errors due to a lack of syntactic knowledge. While syntactic information is crucial for EE, there is a need for effective methods to incorporate syntactic knowledge into PLMs. To address this gap, we present a novel method to incorporate syntactic information into PLM-based models for EE, which do not require external syntactic parsers to produce syntactic features of task data. Instead, our proposed soft syntactic reinforcement (SSR) mechanism learns to select syntax-related dimensions of PLM representation during pretraining on a standard dependency corpus. The adapted PLM weights and the syntax-aware representation then facilitate the model’s prediction over the task data. On both sentence-level and document-level EE benchmark datasets, our proposed method achieves state-of-the-art results, outperforming baseline models and existing syntactic reinforcement methods. To the best of our knowledge, this is the first work in this direction. Our code is available at https://github.com/Anran971/sre-naacl25.

We introduce AudioBench, a universal benchmark designed to evaluate Audio Large Language Models (AudioLLMs). It encompasses 8 distinct tasks and 26 datasets, among which, 7 are newly proposed datasets. The evaluation targets three main aspects: speech understanding, audio scene understanding, and voice understanding (paralinguistic). Despite recent advancements, there lacks a comprehensive benchmark for AudioLLMs on instruction following capabilities conditioned on audio signals. AudioBench addresses this gap by setting up datasets as well as desired evaluation metrics. Besides, we also evaluated the capabilities of five popular models and found that no single model excels consistently across all tasks. We outline the research outlook for AudioLLMs and anticipate that our open-sourced evaluation toolkit, data, and leaderboard will offer a robust testbed for future model developments.

Recent speech-LLMs have shown impressive performance in tasks like transcription and translation, yet they remain limited in understanding the paralinguistic aspects of speech crucial for social and emotional intelligence. We propose CP-Bench, a benchmark for evaluating speech-LLMs on contextual paralinguistic reasoning the integration of verbal content with non-verbal cues like emotion and prosody. The benchmark includes two curated question answering (QA) datasets requiring both linguistic and empathetic understanding. We evaluate state-of-the-art speech-LLMs from both open and closed-source models and perform a comprehensive analysis across different question types. The top two models were further analyzed under temperature tuning to understand its effect on this task. Our benchmark reveals a key gap in existing evaluations and offers insights into building more context-aware and emotionally intelligent speech-capable LLMs.

We introduce MERaLiON-AudioLLM, the first general-purpose audio-based large language model designed for multitask learning, with a particular focus on Singlish understanding. Trained on 62 million multimodal instruction samples comprising a total of 260k hours of audio, it exhibits strong generalization across a diverse set of tasks, including—but not limited to—automatic speech recognition, spoken question answering, speech translation, and paralinguistic analysis. Our results show significant improvements in local speech recognition and task-specific understanding, making MERaLiON-AudioLLM a leading solution for region-specific AI applications. An interactive demo has been developed to enable user-friendly interactions, supported by a backend with customized caching and load-balancing mechanisms. We benchmark the model across a broad range of multilingual and multitask scenarios, where it demonstrates competitive performance compared to other open-source models. The demo page, model weights and videos are publically accessible.