HeadInfer: Memory-Efficient LLM Inference by Head-wise Offloading

Transformer-based large language models (LLMs) demonstrate impressive performance in long context generation. Extending the context length has disproportionately shifted the memory footprint of LLMs during inference to the key-value cache (KV cache). In this paper, we propose HEADINFER, which offloads the KV cache to CPU RAM while avoiding the need to fully store the KV cache for any transformer layer on the GPU. HEADINFER employs a fine-grained, head-wise offloading strategy, maintaining only selective attention heads KV cache on the GPU while computing attention output dynamically. Through roofline analysis, we demonstrate that HEADINFER maintains computational efficiency while significantly reducing memory footprint. We evaluate HEADINFER on the Llama-3-8B model with a 1-million-token sequence, reducing the GPU memory footprint of the KV cache from 128 GB to 1 GB and the total GPU memory usage from 207 GB to 17 GB, achieving a 92% reduction compared to BF16 baseline inference. Notably, HEADINFER enables 4-million-token inference with an 8B model on a single consumer GPU with 24GB memory (e.g., NVIDIA RTX 4090) without approximation methods.

Probing LLM World Models: Enhancing Guesstimation with Wisdom of Crowds Decoding

Guesstimation, the task of making approximate quantity estimates, is a common real-world challenge. However, it has been largely overlooked in large language models (LLMs) and vision language models (VLMs) research. We introduce a novel guesstimation dataset, MARBLES. This dataset requires one to estimate how many items (e.g., marbles) can fit into containers (e.g., a one-cup measuring cup), both with and without accompanying images. Inspired by the social science concept of the ``Wisdom of Crowds'' (WOC) - taking the median from estimates from a crowd), which has proven effective in guesstimation, we propose ``WOC decoding'' strategy for LLM guesstimation. We show that LLMs/VLMs perform well on guesstimation, suggesting that they possess some level of a "world model" necessary for guesstimation. Moreover, similar to human performance, the WOC decoding method improves LLM/VLM guesstimation accuracy. Furthermore, the inclusion of images in the multimodal condition enhances model performance. These results highlight the value of WOC decoding strategy for LLMs/VLMs and position guesstimation as a probe for evaluating LLMs/VLMs' world model. As LLMs' world model is a fundamental prerequisite for many real-world tasks, e.g., human-AI teaming, our findings have broad implications for the AI community.

No Preference Left Behind: Group Distributional Preference Optimization

Preferences within a group of people are not uniform but follow a distribution. While existing alignment methods like Direct Preference Optimization (DPO) attempt to steer models to reflect human preferences, they struggle to capture the distributional pluralistic preferences within a group. These methods often skew toward dominant preferences, overlooking the diversity of opinions, especially when conflicting preferences arise. To address this issue, we propose Group Distribution Preference Optimization (GDPO), a novel framework that aligns language models with the distribution of preferences within a group by incorporating the concept of beliefs that shape individual preferences. GDPO calibrates a language model using statistical estimation of the group's belief distribution and aligns the model with belief-conditioned preferences, offering a more inclusive alignment framework than traditional methods. In experiments using both synthetic controllable opinion generation and real-world movie review datasets, we show that DPO fails to align with the targeted belief distributions, while GDPO consistently reduces this alignment gap during training. Moreover, our evaluation metrics demonstrate that GDPO outperforms existing approaches in aligning with group distributional preferences, marking a significant advance in pluralistic alignment.

P3S-Diffusion:A Selective Subject-driven Generation Framework via Point Supervision

Recent research in subject-driven generation increasingly emphasizes the importance of selective subject features. Nevertheless, accurately selecting the content in a given reference image still poses challenges, especially when selecting the similar subjects in an image (e.g., two different dogs). Some methods attempt to use text prompts or pixel masks to isolate specific elements. However, text prompts often fall short in precisely describing specific content, and pixel masks are often expensive. To address this, we introduce P3S-Diffusion, a novel architecture designed for context-selected subject-driven generation via point supervision. P3S-Diffusion leverages minimal cost label (e.g., points) to generate subject-driven images. During fine-tuning, it can generate an expanded base mask from these points, obviating the need for additional segmentation models. The mask is employed for inpainting and aligning with subject representation. The P3S-Diffusion preserves fine features of the subjects through Multi-layers Condition Injection. Enhanced by the Attention Consistency Loss for improved training, extensive experiments demonstrate its excellent feature preservation and image generation capabilities.

Improving Bilingual Capabilities of Language Models to Support Diverse Linguistic Practices in Education

Large language models (LLMs) offer promise in generating educational content, providing instructor feedback, and reducing teacher workload on assessments. While prior studies have focused on studying LLM-powered learning analytics, limited research has examined how effective LLMs are in a bilingual context. In this paper, we study the effectiveness of multilingual large language models (MLLMs) across monolingual (English-only, Spanish-only) and bilingual (Spanglish) student writing. We present a learning analytics use case that details LLM performance in assessing acceptable and unacceptable explanations of Science and Social Science concepts. Our findings reveal a significant bias in the grading performance of pre-trained models for bilingual writing compared to English-only and Spanish-only writing. Following this, we fine-tune open-source MLLMs including Llama 3.1 and Mistral NeMo using synthetic datasets generated in English, Spanish, and Spanglish. Our experiments indicate that the models perform significantly better for all three languages after fine-tuning with bilingual data. This study highlights the potential of enhancing MLLM effectiveness to support authentic language practices amongst bilingual learners. It also aims to illustrate the value of incorporating non-English languages into the design and implementation of language models in education.

Large Language Models Are Active Critics in NLG Evaluation

The conventional paradigm of using large language models (LLMs) for evaluating natural language generation (NLG) systems typically relies on two key inputs: (1) a clear definition of the NLG task to be evaluated and (2) a list of pre-defined evaluation criteria. This process treats LLMs as ''passive critics,'' strictly following human-defined criteria for evaluation. However, as new NLG tasks emerge, the criteria for assessing text quality can vary greatly. Consequently, these rigid evaluation methods struggle to adapt to diverse NLG tasks without extensive prompt engineering customized for each specific task. To address this limitation, we introduce Active-Critic, a novel LLM-based NLG evaluation protocol that enables LLMs to function as ''active critics.'' Specifically, our protocol comprises two key stages. In the first stage, the LLM is instructed to infer the target NLG task and establish relevant evaluation criteria from the data. Building on this self-inferred information, the second stage dynamically optimizes the prompt to guide the LLM toward more human-aligned scoring decisions, while also generating detailed explanations to justify its evaluations. Experiments across four NLG evaluation tasks show that our approach achieves stronger alignment with human judgments than state-of-the-art evaluation methods. Our comprehensive analysis further highlights the effectiveness and explainability of Active-Critic with only a small amount of labeled data. We will share our code and data on GitHub.

COMMA: A Communicative Multimodal Multi-Agent Benchmark

The rapid advances of multi-modal agents built on large foundation models have largely overlooked their potential for language-based communication between agents in collaborative tasks. This oversight presents a critical gap in understanding their effectiveness in real-world deployments, particularly when communicating with humans. Existing agentic benchmarks fail to address key aspects of inter-agent communication and collaboration, particularly in scenarios where agents have unequal access to information and must work together to achieve tasks beyond the scope of individual capabilities. To fill this gap, we introduce a novel benchmark designed to evaluate the collaborative performance of multimodal multi-agent systems through language communication. Our benchmark features a variety of scenarios, providing a comprehensive evaluation across four key categories of agentic capability in a communicative collaboration setting. By testing both agent-agent and agent-human collaborations using open-source and closed-source models, our findings reveal surprising weaknesses in state-of-the-art models, including proprietary models like GPT-4o. These models struggle to outperform even a simple random agent baseline in agent-agent collaboration and only surpass the random baseline when a human is involved.

From Challenges and Pitfalls to Recommendations and Opportunities: Implementing Federated Learning in Healthcare

Federated learning holds great potential for enabling large-scale healthcare research and collaboration across multiple centres while ensuring data privacy and security are not compromised. Although numerous recent studies suggest or utilize federated learning based methods in healthcare, it remains unclear which ones have potential clinical utility. This review paper considers and analyzes the most recent studies up to May 2024 that describe federated learning based methods in healthcare. After a thorough review, we find that the vast majority are not appropriate for clinical use due to their methodological flaws and/or underlying biases which include but are not limited to privacy concerns, generalization issues, and communication costs. As a result, the effectiveness of federated learning in healthcare is significantly compromised. To overcome these challenges, we provide recommendations and promising opportunities that might be implemented to resolve these problems and improve the quality of model development in federated learning with healthcare.

Streamlining Forest Wildfire Surveillance: AI-Enhanced UAVs Utilizing the FLAME Aerial Video Dataset for Lightweight and Efficient Monitoring

In recent years, unmanned aerial vehicles (UAVs) have played an increasingly crucial role in supporting disaster emergency response efforts by analyzing aerial images. While current deep-learning models focus on improving accuracy, they often overlook the limited computing resources of UAVs. This study recognizes the imperative for real-time data processing in disaster response scenarios and introduces a lightweight and efficient approach for aerial video understanding. Our methodology identifies redundant portions within the video through policy networks and eliminates this excess information using frame compression techniques. Additionally, we introduced the concept of a `station point,' which leverages future information in the sequential policy network, thereby enhancing accuracy. To validate our method, we employed the wildfire FLAME dataset. Compared to the baseline, our approach reduces computation costs by more than 13 times while boosting accuracy by 3$\%$. Moreover, our method can intelligently select salient frames from the video, refining the dataset. This feature enables sophisticated models to be effectively trained on a smaller dataset, significantly reducing the time spent during the training process.

How Well Do LLMs Identify Cultural Unity in Diversity?

Much work on the cultural awareness of large language models (LLMs) focuses on the models' sensitivity to geo-cultural diversity. However, in addition to cross-cultural differences, there also exists common ground across cultures. For instance, a bridal veil in the United States plays a similar cultural-relevant role as a honggaitou in China. In this study, we introduce a benchmark dataset CUNIT for evaluating decoder-only LLMs in understanding the cultural unity of concepts. Specifically, CUNIT consists of 1,425 evaluation examples building upon 285 traditional cultural-specific concepts across 10 countries. Based on a systematic manual annotation of cultural-relevant features per concept, we calculate the cultural association between any pair of cross-cultural concepts. Built upon this dataset, we design a contrastive matching task to evaluate the LLMs' capability to identify highly associated cross-cultural concept pairs. We evaluate 3 strong LLMs, using 3 popular prompting strategies, under the settings of either giving all extracted concept features or no features at all on CUNIT Interestingly, we find that cultural associations across countries regarding clothing concepts largely differ from food. Our analysis shows that LLMs are still limited to capturing cross-cultural associations between concepts compared to humans. Moreover, geo-cultural proximity shows a weak influence on model performance in capturing cross-cultural associations.