Vision-Language Models (VLMs) have demonstrated impressive performance across a versatile set of tasks. A key challenge in accelerating VLMs is storing and accessing the large Key-Value (KV) cache that encodes long visual contexts, such as images or videos. While existing KV cache compression methods are effective for Large Language Models (LLMs), directly migrating them to VLMs yields suboptimal accuracy and speedup. To bridge the gap, we propose VL-Cache, a novel KV cache compression recipe tailored for accelerating VLM inference. In this paper, we first investigate the unique sparsity pattern of VLM attention by distinguishing visual and text tokens in prefill and decoding phases. Based on these observations, we introduce a layer-adaptive sparsity-aware cache budget allocation method that effectively distributes the limited cache budget across different layers, further reducing KV cache size without compromising accuracy. Additionally, we develop a modality-aware token scoring policy to better evaluate the token importance. Empirical results on multiple benchmark datasets demonstrate that retaining only 10% of KV cache achieves accuracy comparable to that with full cache. In a speed benchmark, our method accelerates end-to-end latency of generating 100 tokens by up to 2.33x and speeds up decoding by up to 7.08x, while reducing the memory footprint of KV cache in GPU by 90%.

Open-source Large Language Models (LLMs) have recently demonstrated remarkable capabilities in natural language understanding and generation, leading to widespread adoption across various domains. However, their increasing model sizes render local deployment impractical for individual users, pushing many to rely on computing service providers for inference through a blackbox API. This reliance introduces a new risk: a computing provider may stealthily substitute the requested LLM with a smaller, less capable model without consent from users, thereby delivering inferior outputs while benefiting from cost savings. Existing verifiable computing solutions based on cryptographic or game-theoretic techniques are either computationally uneconomical or rest on strong assumptions. By training a proxy task on these outputs and requiring the computing provider to return both the generated text and the processed intermediate outputs, users can reliably verify whether the computing provider is acting honestly. In addition, the integration of a secret mechanism further enhances the security of our protocol. We thoroughly analyze our protocol under multiple strong and adaptive adversarial scenarios. In recent years, Large Language Models (LLMs) have achieved unprecedented success across a broad array of tasks and domains (Achiam et al., 2023; Dubey et al., 2024; Yang et al., 2024). Alongside this progress, open-source LLMs have proliferated, offering increasingly sophisticated and capable models to the broader research community (Touvron et al., 2023b; Black et al., 2022; Le Scao et al., 2023; Jiang et al., 2023; Almazrouei et al., 2023; Zhang et al., 2023). Many of these open-source LLMs now rival, or even surpass, their closed-source counterparts in performance (Chiang et al., 2023; Almazrouei et al., 2023; Dubey et al., 2024), while remaining freely accessible. However, as model capacity grows, it typically comes with a corresponding increase in the number of model parameters, which directly drives up the computational demands, particularly in terms of memory and processing power (Kukreja et al., 2024).

In recent years, there has been a trend in the field of Reinforcement Learning (RL) towards large action models trained offline on large-scale datasets via sequence modeling. Existing models are primarily based on the Transformer architecture, which result in powerful agents. However, due to slow inference times, Transformer-based approaches are impractical for real-time applications, such as robotics. Recently, modern recurrent architectures, such as xLSTM and Mamba, have been proposed that exhibit parallelization benefits during training similar to the Transformer architecture while offering fast inference. In this work, we study the aptitude of these modern recurrent architectures for large action models. Consequently, we propose a Large Recurrent Action Model (LRAM) with an xLSTM at its core that comes with linear-time inference complexity and natural sequence length extrapolation abilities. Experiments on 432 tasks from 6 domains show that LRAM compares favorably to Transformers in terms of performance and speed.

Artificial intelligence agents are increasingly capable of performing complex tasks across various domains. As these agents advance, there is a growing need to accurately measure and benchmark their capabilities, particularly in accelerating AI research and development. Current benchmarks focus on general machine learning tasks, but lack comprehensive evaluation methods for assessing AI agents' abilities in tackling research-level problems and competition-level challenges in the field of AI. We present the ML Research Benchmark (MLRB), comprising 7 competition-level tasks derived from recent machine learning conference tracks. These tasks span activities typically undertaken by AI researchers, including model training efficiency, pretraining on limited data, domain specific fine-tuning, and model compression. This paper introduces a novel benchmark and evaluates it using agent scaffolds powered by frontier models, including Claude-3 and GPT-4o. The results indicate that the Claude-3.5 Sonnet agent performs best across our benchmark, excelling in planning and developing machine learning models. However, both tested agents struggled to perform non-trivial research iterations. We observed significant performance variations across tasks, highlighting the complexity of AI development and the challenges in creating versatile agent scaffolds. While current AI agents can successfully navigate complex instructions and produce baseline results, they fall short of the capabilities required for advanced AI research. The ML Research Benchmark provides a valuable framework for assessing and comparing AI agents on tasks mirroring real-world AI research challenges.

We introduce an innovative approach to advancing semantic understanding in zero-shot object goal navigation (ZS-OGN), enhancing the autonomy of robots in unfamiliar environments. Traditional reliance on labeled data has been a limitation for robotic adaptability, which we address by employing a dual-component framework that integrates a GLIP Vision Language Model for initial detection and an Instruction-BLIP model for validation. This combination not only refines object and environmental recognition but also fortifies the semantic interpretation, pivotal for navigational decision-making. Our method, rigorously tested in both simulated and real-world settings, exhibits marked improvements in navigation precision and reliability.

Transformer-based autoregressive sampling has been the major bottleneck for slowing down large language model inferences. One effective way to accelerate inference is \emph{Speculative Decoding}, which employs a small model to sample a sequence of draft tokens and a large model to validate. Given its empirical effectiveness, the theoretical understanding of Speculative Decoding is falling behind. This paper tackles this gap by conceptualizing the decoding problem via markov chain abstraction and studying the key properties, \emph{output quality and inference acceleration}, from a theoretical perspective. Our analysis covers the theoretical limits of speculative decoding, batch algorithms, and output quality-inference acceleration tradeoffs. Our results reveal the fundamental connections between different components of LLMs via total variation distances and show how they jointly affect the efficiency of decoding algorithms.

Generating music from images can enhance various applications, including background music for photo slideshows, social media experiences, and video creation. This paper presents an emotion-guided image-to-music generation framework that leverages the Valence-Arousal (VA) emotional space to produce music that aligns with the emotional tone of a given image. Unlike previous models that rely on contrastive learning for emotional consistency, the proposed approach directly integrates a VA loss function to enable accurate emotional alignment. The model employs a CNN-Transformer architecture, featuring pre-trained CNN image feature extractors and three Transformer encoders to capture complex, high-level emotional features from MIDI music. Three Transformer decoders refine these features to generate musically and emotionally consistent MIDI sequences. Experimental results on a newly curated emotionally paired image-MIDI dataset demonstrate the proposed model's superior performance across metrics such as Polyphony Rate, Pitch Entropy, Groove Consistency, and loss convergence.

Developing algorithms to differentiate between machine-generated texts and human-written texts has garnered substantial attention in recent years. Existing methods in this direction typically concern an offline setting where a dataset containing a mix of real and machine-generated texts is given upfront, and the task is to determine whether each sample in the dataset is from a large language model (LLM) or a human. However, in many practical scenarios, sources such as news websites, social media accounts, or on other forums publish content in a streaming fashion. Therefore, in this online scenario, how to quickly and accurately determine whether the source is an LLM with strong statistical guarantees is crucial for these media or platforms to function effectively and prevent the spread of misinformation and other potential misuse of LLMs. To tackle the problem of online detection, we develop an algorithm based on the techniques of sequential hypothesis testing by betting that not only builds upon and complements existing offline detection techniques but also enjoys statistical guarantees, which include a controlled false positive rate and the expected time to correctly identify a source as an LLM. Experiments were conducted to demonstrate the effectiveness of our method.

We seek to address a core challenge facing current Large Language Models (LLMs). LLMs have demonstrated superior performance in many tasks, yet continue to struggle with reasoning problems on explicit graphs that require multiple steps. To address this gap, we introduce a novel benchmark designed to evaluate LLM performance on classical algorithmic reasoning tasks on explicit graphs. Our benchmark encompasses five fundamental algorithms: Breadth-First Search (BFS) and Depth-First Search (DFS) for connectivity, Dijkstra's algorithm and Floyd-Warshall algorithm for all nodes shortest path, and Prim's Minimum Spanning Tree (MST-Prim's) algorithm. Through extensive experimentation, we assess the capabilities of state-of-the-art LLMs in executing these algorithms step-by-step and systematically evaluate their performance at each stage. Our findings highlight the persistent challenges LLMs face in this domain and underscore the necessity for advanced prompting techniques and algorithmic instruction to enhance their graph reasoning abilities. This work presents MAGMA, the first comprehensive benchmark focused on LLMs completing classical graph algorithms, and provides a critical step toward understanding and improving their structured problem-solving skills.

Property inheritance -- a phenomenon where novel properties are projected from higher level categories (e.g., birds) to lower level ones (e.g., sparrows) -- provides a unique window into how humans organize and deploy conceptual knowledge. It is debated whether this ability arises due to explicitly stored taxonomic knowledge vs. simple computations of similarity between mental representations. How are these mechanistic hypotheses manifested in contemporary language models? In this work, we investigate how LMs perform property inheritance with behavioral and causal representational analysis experiments. We find that taxonomy and categorical similarities are not mutually exclusive in LMs' property inheritance behavior. That is, LMs are more likely to project novel properties from one category to the other when they are taxonomically related and at the same time, highly similar. Our findings provide insight into the conceptual structure of language models and may suggest new psycholinguistic experiments for human subjects.