MSCCL++: Rethinking GPU Communication Abstractions for Cutting-edge AI Applications

Modern cutting-edge AI applications are being developed over fast-evolving, heterogeneous, nascent hardware devices. This requires frequent reworking of the AI software stack to adopt bottom-up changes from new hardware, which takes time for general-purpose software libraries. Consequently, real applications often develop custom software stacks optimized for their specific workloads and hardware. Custom stacks help quick development and optimization, but incur a lot of redundant efforts across applications in writing non-portable code. This paper discusses an alternative communication library interface for AI applications that offers both portability and performance by reducing redundant efforts while maintaining flexibility for customization. We present MSCCL++, a novel abstraction of GPU communication based on separation of concerns: (1) a primitive interface provides a minimal hardware abstraction as a common ground for software and hardware developers to write custom communication, and (2) higher-level portable interfaces and specialized implementations enable optimization for different hardware environments. This approach makes the primitive interface reusable across applications while enabling highly flexible optimization. Compared to state-of-the-art baselines (NCCL, RCCL, and MSCCL), MSCCL++ achieves speedups of up to 3.8$\times$ for collective communication and up to 15\% for real-world AI inference workloads. MSCCL++ is in production of multiple AI services provided by Microsoft Azure, and is also adopted by RCCL, the GPU collective communication library maintained by AMD. MSCCL++ is open-source and available at https://github.com/microsoft/mscclpp.

WHERE and WHICH: Iterative Debate for Biomedical Synthetic Data Augmentation

In Biomedical Natural Language Processing (BioNLP) tasks, such as Relation Extraction, Named Entity Recognition, and Text Classification, the scarcity of high-quality data remains a significant challenge. This limitation poisons large language models to correctly understand relationships between biological entities, such as molecules and diseases, or drug interactions, and further results in potential misinterpretation of biomedical documents. To address this issue, current approaches generally adopt the Synthetic Data Augmentation method which involves similarity computation followed by word replacement, but counterfactual data are usually generated. As a result, these methods disrupt meaningful word sets or produce sentences with meanings that deviate substantially from the original context, rendering them ineffective in improving model performance. To this end, this paper proposes a biomedical-dedicated rationale-based synthetic data augmentation method. Beyond the naive lexicon similarity, specific bio-relation similarity is measured to hold the augmented instance having a strong correlation with bio-relation instead of simply increasing the diversity of augmented data. Moreover, a multi-agents-involved reflection mechanism helps the model iteratively distinguish different usage of similar entities to escape falling into the mis-replace trap. We evaluate our method on the BLURB and BigBIO benchmark, which includes 9 common datasets spanning four major BioNLP tasks. Our experimental results demonstrate consistent performance improvements across all tasks, highlighting the effectiveness of our approach in addressing the challenges associated with data scarcity and enhancing the overall performance of biomedical NLP models.

FReM: A Flexible Reasoning Mechanism for Balancing Quick and Slow Thinking in Long-Context Question Answering

Long-context question-answering (LCQA) systems have greatly benefited from the powerful reasoning capabilities of large language models (LLMs), which can be categorized into slow and quick reasoning modes. However, both modes have their limitations. Slow thinking generally leans to explore every possible reasoning path, which leads to heavy overthinking and wastes time. Quick thinking usually relies on pattern matching rather than truly understanding the query logic, which misses proper understanding. To address these issues, we propose FReM: Flexible Reasoning Mechanism, a method that adjusts reasoning depth according to the complexity of each question. Specifically, FReM leverages synthetic reference QA examples to provide an explicit chain of thought, enabling efficient handling of simple queries while allowing deeper reasoning for more complex ones. By doing so, FReM helps quick-thinking models move beyond superficial pattern matching and narrows the reasoning space for slow-thinking models to avoid unnecessary exploration. Experiments on seven QA datasets show that FReM improves reasoning accuracy and scalability, particularly for complex multihop questions, indicating its potential to advance LCQA methodologies.

EventWeave: A Dynamic Framework for Capturing Core and Supporting Events in Dialogue Systems

Existing large language models (LLMs) have shown remarkable progress in dialogue systems. However, many approaches still overlook the fundamental role of events throughout multi-turn interactions, leading to \textbf{incomplete context tracking}. Without tracking these events, dialogue systems often lose coherence and miss subtle shifts in user intent, causing disjointed responses. To bridge this gap, we present \textbf{EventWeave}, an event-centric framework that identifies and updates both core and supporting events as the conversation unfolds. Specifically, we organize these events into a dynamic event graph, which represents the interplay between \textbf{core events} that shape the primary idea and \textbf{supporting events} that provide critical context during the whole dialogue. By leveraging this dynamic graph, EventWeave helps models focus on the most relevant events when generating responses, thus avoiding repeated visits of the entire dialogue history. Experimental results on two benchmark datasets show that EventWeave improves response quality and event relevance without fine-tuning.

Sigma: Differential Rescaling of Query, Key and Value for Efficient Language Models

We introduce Sigma, an efficient large language model specialized for the system domain, empowered by a novel architecture including DiffQKV attention, and pre-trained on our meticulously collected system domain data. DiffQKV attention significantly enhances the inference efficiency of Sigma by optimizing the Query (Q), Key (K), and Value (V) components in the attention mechanism differentially, based on their varying impacts on the model performance and efficiency indicators. Specifically, we (1) conduct extensive experiments that demonstrate the model's varying sensitivity to the compression of K and V components, leading to the development of differentially compressed KV, and (2) propose augmented Q to expand the Q head dimension, which enhances the model's representation capacity with minimal impacts on the inference speed. Rigorous theoretical and empirical analyses reveal that DiffQKV attention significantly enhances efficiency, achieving up to a 33.36% improvement in inference speed over the conventional grouped-query attention (GQA) in long-context scenarios. We pre-train Sigma on 6T tokens from various sources, including 19.5B system domain data that we carefully collect and 1T tokens of synthesized and rewritten data. In general domains, Sigma achieves comparable performance to other state-of-arts models. In the system domain, we introduce the first comprehensive benchmark AIMicius, where Sigma demonstrates remarkable performance across all tasks, significantly outperforming GPT-4 with an absolute improvement up to 52.5%.

GODIVA: Generating Open-DomaIn Videos from nAtural Descriptions

Generating videos from text is a challenging task due to its high computational requirements for training and infinite possible answers for evaluation. Existing works typically experiment on simple or small datasets, where the generalization ability is quite limited. In this work, we propose GODIVA, an open-domain text-to-video pretrained model that can generate videos from text in an auto-regressive manner using a three-dimensional sparse attention mechanism. We pretrain our model on Howto100M, a large-scale text-video dataset that contains more than 136 million text-video pairs. Experiments show that GODIVA not only can be fine-tuned on downstream video generation tasks, but also has a good zero-shot capability on unseen texts. We also propose a new metric called Relative Matching (RM) to automatically evaluate the video generation quality. Several challenges are listed and discussed as future work.

CHIME: Cross-passage Hierarchical Memory Network for Generative Review Question Answering

We introduce CHIME, a cross-passage hierarchical memory network for question answering (QA) via text generation. It extends XLNet introducing an auxiliary memory module consisting of two components: the context memory collecting cross-passage evidences, and the answer memory working as a buffer continually refining the generated answers. Empirically, we show the efficacy of the proposed architecture in the multi-passage generative QA, outperforming the state-of-the-art baselines with better syntactically well-formed answers and increased precision in addressing the questions of the AmazonQA review dataset. An additional qualitative analysis revealed the interpretability introduced by the memory module.

Context-aware Embedding for Targeted Aspect-based Sentiment Analysis

Attention-based neural models were employed to detect the different aspects and sentiment polarities of the same target in targeted aspect-based sentiment analysis (TABSA). However, existing methods do not specifically pre-train reasonable embeddings for targets and aspects in TABSA. This may result in targets or aspects having the same vector representations in different contexts and losing the context-dependent information. To address this problem, we propose a novel method to refine the embeddings of targets and aspects. Such pivotal embedding refinement utilizes a sparse coefficient vector to adjust the embeddings of target and aspect from the context. Hence the embeddings of targets and aspects can be refined from the highly correlative words instead of using context-independent or randomly initialized vectors. Experiment results on two benchmark datasets show that our approach yields the state-of-the-art performance in TABSA task.