The quadratic complexity of the attention mechanism and the substantial memory footprint of the Key-Value (KV) cache present severe computational and memory challenges for Large Language Models (LLMs) processing long contexts. Existing retrieval-based methods often compromise semantic integrity through fixed-size chunking and suffer from inefficient linear scanning. In this paper, we propose LycheeCluster, a novel method for efficient KV cache management. LycheeCluster preserves local semantic coherence via boundary-aware chunking and constructs a recursive hierarchical index rooted in the triangle inequality. This design transforms cache retrieval from a linear scan into a theoretically bounded, logarithmic-time pruning process, while a lazy update strategy supports efficient streaming generation. Experiments demonstrate that LycheeCluster achieves up to a 3.6× end-to-end inference speedup with negligible degradation in model performance, outperforming state-of-the-art KV cache management methods (e.g., Quest, ClusterKV).

Large Language Models (LLMs) are increasingly deployed in settings where rare but severe harmful generations can have significant consequences. Existing Risk-Averse RLHF addresses this issue by optimizing Conditional Value-at-Risk (CVaR), but it trains policies for fixed risk levels and therefore cannot adjust the desired degree of risk aversion at inference time. In this paper, we propose *risk-conditioned RLHF*, a framework that trains a single policy that provides a continuous risk-control interface, enabling users to select different degrees of risk aversion without retraining or deploying multiple risk-specific models. Experiments across multiple benchmarks demonstrate that a single risk-conditioned policy can adapt to different risk levels at inference time, enabling more flexible and risk-aware LLM deployment. This paper contains example data that may be offensive or harmful.

Summarizing biomedical discovery from genomics data using natural languages is an essential step in biomedical research but is mostly done manually. Here, we introduce Textomics, a novel dataset of genomics data description, which contains 22,273 pairs of genomics data matrices and their summaries. Each summary is written by the researchers who generated the data and associated with a scientific paper. Based on this dataset, we study two novel tasks: generating textual summary from a genomics data matrix and vice versa. Inspired by the successful applications of k nearest neighbors in modeling genomics data, we propose a kNN-Vec2Text model to address these tasks and observe substantial improvement on our dataset. We further illustrate how Textomics can be used to advance other applications, including evaluating scientific paper embeddings and generating masked templates for scientific paper understanding. Textomics serves as the first benchmark for generating textual summaries for genomics data and we envision it will be broadly applied to other biomedical and natural language processing applications.

Images are more than a collection of objects or attributes — they represent a web of relationships among interconnected objects. Scene Graph has emerged as a new modality as a structured graphical representation of images. Scene Graph encodes objects as nodes connected via pairwise relations as edges. To support question answering on scene graphs, we propose GraphVQA, a language-guided graph neural network framework that translates and executes a natural language question as multiple iterations of message passing among graph nodes. We explore the design space of GraphVQA framework, and discuss the trade-off of different design choices. Our experiments on GQA dataset show that GraphVQA outperforms the state-of-the-art accuracy by a large margin (88.43% vs. 94.78%).