Large Audio Language Models have shown impressive performance on single-clip audio language tasks such as automatic speech recognition, captioning, and sound event recognition. Yet, their ability to reason over interleaved multi-audio contexts-where answering a query requires relating information across multiple audio clips-remains limited. We present PolyAudio, a LALM built on Audio Flamingo 3 that targets multi-audio understanding via instruction tuning rather than massive-scale pre-training, and PolyAudio-Instruct, a high-quality instruction-tuning dataset consisting of 1.3M+ QA pairs, spanning over 14 task subsets to empower multi-audio understanding and reasoning. PolyAudio uses an explicit interleaved representation with clip indexing to encourage faithful grounding and reduce ambiguity in multi-clip references. We evaluate PolyAudio on a diverse suite of multi-audio benchmarks alongside standard single-audio tasks. PolyAudio achieves strong performance on multi-audio reasoning, outperforming competitive baselines that are also often limited to reasoning over up-to 2 audio clips, while preserving robust single-clip performance. Overall, our results suggest that precise, academic-scale multi-audio instruction tuning can unlock advanced cross-clip reasoning capabilities, enabling more capable audio-centric assistants.

Multimodal large language models (MLLMs) have made substantial advancements in video understanding, yet the reliability of their responses remains underexplored. This work presents a diagnostic study of absent answer detection for MLLMs in video understanding, where the correct answer is deliberately excluded from the candidate set and a reliable model is expected to recognize that no valid option exists. We evaluate the absent answer detection behavior under three settings: multiple-choice questions augmented with a “None of the Above” option, open-ended generation with a detection instruction, and standard evaluation without any guidance. Across a diverse set of models and benchmarks, we find that MLLMs overwhelmingly select plausible distractors rather than detecting the absent answer. This failure is more pronounced in temporal reasoning tasks and worsens with denser frame sampling. We further explore chain-of-thought prompting as a mitigation strategy and find that while it substantially improves detection rates, performance remains unsatisfactory, suggesting that prompting-based strategies alone are insufficient to fully address this limitation. These findings expose a systematic failure in absent answer detection and highlight the need for explicit detection mechanisms in multimodal systems.

Safety evaluation of multimodal large language models requires tracking not only whether an attack succeeds, but also how the interaction unfolds across turns and input modalities. We present MUSE (Multimodal Unified Safety Evaluation), an open-source, browser-based, run-centric platform for multimodal safety evaluation. MUSE treats each attack run as the persistent unit of execution, inspection, and analysis, preserving its configuration, multi-turn trajectory, delivered modalities and media, target responses, and safety judgments. A five-level response taxonomy further distinguishes full Compliance from Partial Compliance and refusal behavior, yielding hard ASR, soft ASR, and gray-zone width (GZW).

Across 11,700 evaluations on six multimodal LLMs, direct text-only requests yield only 3.1% macro hard ASR and 4.4% soft ASR, while iterative attack procedures are substantially more effective. Attack effectiveness also varies substantially with the attacker backbone. In contrast, Inter-Turn Modality Switching (ITMS), evaluated as a controlled delivery-modality probe, does not consistently increase attack success. These results demonstrate the value of run-centric, fine-grained evaluation for characterizing multimodal safety behavior beyond a single binary success metric.

Personalized AI assistants in collaborative settings must balance private user context with shared team knowledge; i.e. the same question can have different answers, depending on who is asking. However, retrieving and reasoning over a growing shared archive, as agents do today, increases latency and cost. We present Goldilocks, a multimodal assistive-agent demo that explores a tiered personal memory architecture which combines a lightweight private working resident memory alongside a shared team memory, both grounded in a common eventstore. A sufficiency gate first tries to answer from the low-cost resident tier and only escalates other tiers when needed. To evaluate long-lived personalized collaborative assistants, we also introduce WhoQA, a 500-question evaluation set built from public GitHub history, where work-related questions and answers trace to real pull requests and commits. On WhoQA, Goldilocks’s deployed small-model configuration beats a strong hybrid RAG baseline by 7.0 points (59.0 vs. 52.0) at 13% lower estimated energy (a tier-weighted token cost), and reaches 62.4 (+10.4) with the answer model held fixed. The system serves 42.6% of queries from the resident tier. A compact social-memory scoping further lifts accuracy on teammate-specific personal details from 53 to 80. While we demonstrate with a set tier and scoping configuration, the system is generalizable to multiple tiers depending on the use case. The live demo lets users explore memory tiers, gating decisions, and interactively compare personalized responses across team personas.

Knowledge Distillation (KD) is one of the most effective approaches to deploying large-scale pre-trained language models in low-latency environments by transferring the knowledge contained in the large-scale models to smaller student models. Prior KD approaches use the soft labels and intermediate activations generated by the teacher to transfer knowledge to the student model parameters alone. In this paper, we show that having access to non-parametric memory in the form of a knowledge base with the teacher’s soft labels and predictions can further improve student generalization. To enable the student to retrieve from the knowledge base effectively, we propose a new framework and loss function that preserves the semantic similarities of teacher and student training examples. We show through extensive experiments that our retrieval mechanism can achieve state-of-the-art performance for task-specific knowledge distillation on the GLUE benchmark.

We introduce MTG, a new benchmark suite for training and evaluating multilingual text generation. It is the first-proposed multilingual multiway text generation dataset with the largest human-annotated data (400k). It includes four generation tasks (story generation, question generation, title generation and text summarization) across five languages (English, German, French, Spanish and Chinese). The multiway setup enables testing knowledge transfer capabilities for a model across languages and tasks. Using MTG, we train and analyze several popular multilingual generation models from different aspects. Our benchmark suite fosters model performance enhancement with more human-annotated parallel data. It provides comprehensive evaluations with diverse generation scenarios. Code and data are available at https://github.com/zide05/MTG.

Despite recent advancements in automated speech recognition (ASR) technologies, reports of unequal performance across speakers of different demographic groups abound. At the same time, the focus on performance metrics such as the Word Error Rate (WER) in prior studies limit the specificity and scope of recommendations that can be offered for system engineering to overcome these challenges. The current study bridges this gap by investigating the performance of Otter’s automatic captioning system on native and non-native English speakers of different language background through a linguistic analysis of segment-level errors. By examining language-specific error profiles for vowels and consonants motivated by linguistic theory, we find that certain categories of errors can be predicted from the phonological structure of a speaker’s native language.

With the continuous upgrading of the summarization systems driven by deep neural networks, researchers have higher requirements on the quality of the generated summaries, which should be not only fluent and informative but also factually correct. As a result, the field of factual evaluation has developed rapidly recently. Despite its initial progress in evaluating generated summaries, the meta-evaluation methodologies of factuality metrics are limited in their opacity, leading to the insufficient understanding of factuality metrics’ relative advantages and their applicability. In this paper, we present an adversarial meta-evaluation methodology that allows us to (i) diagnose the fine-grained strengths and weaknesses of 6 existing top-performing metrics over 24 diagnostic test datasets, (ii) search for directions for further improvement by data augmentation. Our observations from this work motivate us to propose several calls for future research. We make all codes, diagnostic test datasets, trained factuality models available: https://github.com/zide05/AdvFact.

Neural network-based models augmented with unsupervised pre-trained knowledge have achieved impressive performance on text summarization. However, most existing evaluation methods are limited to an in-domain setting, where summarizers are trained and evaluated on the same dataset. We argue that this approach can narrow our understanding of the generalization ability for different summarization systems. In this paper, we perform an in-depth analysis of characteristics of different datasets and investigate the performance of different summarization models under a cross-dataset setting, in which a summarizer trained on one corpus will be evaluated on a range of out-of-domain corpora. A comprehensive study of 11 representative summarization systems on 5 datasets from different domains reveals the effect of model architectures and generation ways (i.e. abstractive and extractive) on model generalization ability. Further, experimental results shed light on the limitations of existing summarizers. Brief introduction and supplementary code can be found in https://github.com/zide05/CDEvalSumm.

This paper creates a paradigm shift with regard to the way we build neural extractive summarization systems. Instead of following the commonly used framework of extracting sentences individually and modeling the relationship between sentences, we formulate the extractive summarization task as a semantic text matching problem, in which a source document and candidate summaries will be (extracted from the original text) matched in a semantic space. Notably, this paradigm shift to semantic matching framework is well-grounded in our comprehensive analysis of the inherent gap between sentence-level and summary-level extractors based on the property of the dataset. Besides, even instantiating the framework with a simple form of a matching model, we have driven the state-of-the-art extractive result on CNN/DailyMail to a new level (44.41 in ROUGE-1). Experiments on the other five datasets also show the effectiveness of the matching framework. We believe the power of this matching-based summarization framework has not been fully exploited. To encourage more instantiations in the future, we have released our codes, processed dataset, as well as generated summaries in https://github.com/maszhongming/MatchSum.