ToolHop: A Query-Driven Benchmark for Evaluating Large Language Models in Multi-Hop Tool Use

Effective evaluation of multi-hop tool use is critical for analyzing the understanding, reasoning, and function-calling capabilities of large language models (LLMs). However, progress has been hindered by a lack of reliable evaluation datasets. To address this, we present ToolHop, a dataset comprising 995 user queries and 3,912 associated tools, specifically designed for rigorous evaluation of multi-hop tool use. ToolHop ensures diverse queries, meaningful interdependencies, locally executable tools, detailed feedback, and verifiable answers through a novel query-driven data construction approach that includes tool creation, document refinement, and code generation. We evaluate 14 LLMs across five model families (i.e., LLaMA3.1, Qwen2.5, Gemini1.5, Claude3.5, and GPT), uncovering significant challenges in handling multi-hop tool-use scenarios. The leading model, GPT-4o, achieves an accuracy of 49.04%, underscoring substantial room for improvement. Further analysis reveals variations in tool-use strategies for various families, offering actionable insights to guide the development of more effective approaches. Code and data can be found in https://huggingface.co/datasets/bytedance-research/ToolHop.

GeAR: Generation Augmented Retrieval

Document retrieval techniques form the foundation for the development of large-scale information systems. The prevailing methodology is to construct a bi-encoder and compute the semantic similarity. However, such scalar similarity is difficult to reflect enough information and impedes our comprehension of the retrieval results. In addition, this computational process mainly emphasizes the global semantics and ignores the fine-grained semantic relationship between the query and the complex text in the document. In this paper, we propose a new method called $\textbf{Ge}$neration $\textbf{A}$ugmented $\textbf{R}$etrieval ($\textbf{GeAR}$) that incorporates well-designed fusion and decoding modules. This enables GeAR to generate the relevant text from documents based on the fused representation of the query and the document, thus learning to "focus on" the fine-grained information. Also when used as a retriever, GeAR does not add any computational burden over bi-encoders. To support the training of the new framework, we have introduced a pipeline to efficiently synthesize high-quality data by utilizing large language models. GeAR exhibits competitive retrieval and localization performance across diverse scenarios and datasets. Moreover, the qualitative analysis and the results generated by GeAR provide novel insights into the interpretation of retrieval results. The code, data, and models will be released after completing technical review to facilitate future research.

Evidential Calibrated Uncertainty-Guided Interactive Segmentation paradigm for Ultrasound Images

Accurate and robust ultrasound image segmentation is critical for computer-aided diagnostic systems. Nevertheless, the inherent challenges of ultrasound imaging, such as blurry boundaries and speckle noise, often cause traditional segmentation methods to struggle with performance. Despite recent advancements in universal image segmentation, such as the Segment Anything Model, existing interactive segmentation methods still suffer from inefficiency and lack of specialization. These methods rely heavily on extensive accurate manual or random sampling prompts for interaction, necessitating numerous prompts and iterations to reach satisfactory performance. In response to this challenge, we propose the Evidential Uncertainty-Guided Interactive Segmentation (EUGIS), an end-to-end, efficient tiered interactive segmentation paradigm based on evidential uncertainty estimation for ultrasound image segmentation. Specifically, EUGIS harnesses evidence-based uncertainty estimation, grounded in Dempster-Shafer theory and Subjective Logic, to gauge the level of uncertainty in the predictions of model for different regions. By prioritizing sampling the high-uncertainty region, our method can effectively simulate the interactive behavior of well-trained radiologists, enhancing the targeted of sampling while reducing the number of prompts and iterations required.Additionally, we propose a trainable calibration mechanism for uncertainty estimation, which can further optimize the boundary between certainty and uncertainty, thereby enhancing the confidence of uncertainty estimation.

Temporal Information Reconstruction and Non-Aligned Residual in Spiking Neural Networks for Speech Classification

Recently, it can be noticed that most models based on spiking neural networks (SNNs) only use a same level temporal resolution to deal with speech classification problems, which makes these models cannot learn the information of input data at different temporal scales. Additionally, owing to the different time lengths of the data before and after the sub-modules of many models, the effective residual connections cannot be applied to optimize the training processes of these models.To solve these problems, on the one hand, we reconstruct the temporal dimension of the audio spectrum to propose a novel method named as Temporal Reconstruction (TR) by referring the hierarchical processing process of the human brain for understanding speech. Then, the reconstructed SNN model with TR can learn the information of input data at different temporal scales and model more comprehensive semantic information from audio data because it enables the networks to learn the information of input data at different temporal resolutions. On the other hand, we propose the Non-Aligned Residual (NAR) method by analyzing the audio data, which allows the residual connection can be used in two audio data with different time lengths. We have conducted plentiful experiments on the Spiking Speech Commands (SSC), the Spiking Heidelberg Digits (SHD), and the Google Speech Commands v0.02 (GSC) datasets. According to the experiment results, we have achieved the state-of-the-art (SOTA) result 81.02\% on SSC for the test classification accuracy of all SNN models, and we have obtained the SOTA result 96.04\% on SHD for the classification accuracy of all models.

Extract Information from Hybrid Long Documents Leveraging LLMs: A Framework and Dataset

Large Language Models (LLMs) demonstrate exceptional performance in textual understanding and tabular reasoning tasks. However, their ability to comprehend and analyze hybrid text, containing textual and tabular data, remains unexplored. The hybrid text often appears in the form of hybrid long documents (HLDs), which far exceed the token limit of LLMs. Consequently, we apply an Automated Information Extraction framework (AIE) to enable LLMs to process the HLDs and carry out experiments to analyse four important aspects of information extraction from HLDs. Given the findings: 1) The effective way to select and summarize the useful part of a HLD. 2) An easy table serialization way is enough for LLMs to understand tables. 3) The naive AIE has adaptability in many complex scenarios. 4) The useful prompt engineering to enhance LLMs on HLDs. To address the issue of dataset scarcity in HLDs and support future work, we also propose the Financial Reports Numerical Extraction (FINE) dataset. The dataset and code are publicly available in the attachments.

OMG-HD: A High-Resolution AI Weather Model for End-to-End Forecasts from Observations

In recent years, Artificial Intelligence Weather Prediction (AIWP) models have achieved performance comparable to, or even surpassing, traditional Numerical Weather Prediction (NWP) models by leveraging reanalysis data. However, a less-explored approach involves training AIWP models directly on observational data, enhancing computational efficiency and improving forecast accuracy by reducing the uncertainties introduced through data assimilation processes. In this study, we propose OMG-HD, a novel AI-based regional high-resolution weather forecasting model designed to make predictions directly from observational data sources, including surface stations, radar, and satellite, thereby removing the need for operational data assimilation. Our evaluation shows that OMG-HD outperforms both the European Centre for Medium-Range Weather Forecasts (ECMWF)'s high-resolution operational forecasting system, IFS-HRES, and the High-Resolution Rapid Refresh (HRRR) model at lead times of up to 12 hours across the contiguous United States (CONUS) region. We achieve up to a 13% improvement on RMSE for 2-meter temperature, 17% on 10-meter wind speed, 48% on 2-meter specific humidity, and 32% on surface pressure compared to HRRR. Our method shows that it is possible to use AI-driven approaches for rapid weather predictions without relying on NWP-derived weather fields as model input. This is a promising step towards using observational data directly to make operational forecasts with AIWP models.

Prompt Tuning for Item Cold-start Recommendation

The item cold-start problem is crucial for online recommender systems, as the success of the cold-start phase determines whether items can transition into popular ones. Prompt learning, a powerful technique used in natural language processing (NLP) to address zero- or few-shot problems, has been adapted for recommender systems to tackle similar challenges. However, existing methods typically rely on content-based properties or text descriptions for prompting, which we argue may be suboptimal for cold-start recommendations due to 1) semantic gaps with recommender tasks, 2) model bias caused by warm-up items contribute most of the positive feedback to the model, which is the core of the cold-start problem that hinders the recommender quality on cold-start items. We propose to leverage high-value positive feedback, termed pinnacle feedback as prompt information, to simultaneously resolve the above two problems. We experimentally prove that compared to the content description proposed in existing works, the positive feedback is more suitable to serve as prompt information by bridging the semantic gaps. Besides, we propose item-wise personalized prompt networks to encode pinnaclce feedback to relieve the model bias by the positive feedback dominance problem. Extensive experiments on four real-world datasets demonstrate the superiority of our model over state-of-the-art methods. Moreover, PROMO has been successfully deployed on a popular short-video sharing platform, a billion-user scale commercial short-video application, achieving remarkable performance gains across various commercial metrics within cold-start scenarios

WarriorCoder: Learning from Expert Battles to Augment Code Large Language Models

Despite recent progress achieved by code large language models (LLMs), their remarkable abilities are largely dependent on fine-tuning on the high-quality data, posing challenges for data collection and annotation. To address this, current methods often design various data flywheels to gather complex code instructions, enabling models to handle more intricate tasks. However, these approaches typically rely on off-the-shelf datasets and data augmentation from the limited pool of proprietary LLMs (e.g., Claude, GPT4, and so on), which limits the diversity of the constructed data and makes it prone to systemic biases. In this paper, we propose WarriorCoder which learns from expert battles to address these limitations. Specifically, we create an arena for current expert code LLMs, where each model challenges and responds to others' challenges, with evaluations conducted by uninvolved judge models. This competitive framework generates novel training data constructed from scratch, harnessing the strengths of all participants. Experimental results demonstrate that WarriorCoder achieves competitive performance compared to previous methods, even without relying on proprietary LLMs.

Predicting Satisfied User and Machine Ratio for Compressed Images: A Unified Approach

Nowadays, high-quality images are pursued by both humans for better viewing experience and by machines for more accurate visual analysis. However, images are usually compressed before being consumed, decreasing their quality. It is meaningful to predict the perceptual quality of compressed images for both humans and machines, which guides the optimization for compression. In this paper, we propose a unified approach to address this. Specifically, we create a deep learning-based model to predict Satisfied User Ratio (SUR) and Satisfied Machine Ratio (SMR) of compressed images simultaneously. We first pre-train a feature extractor network on a large-scale SMR-annotated dataset with human perception-related quality labels generated by diverse image quality models, which simulates the acquisition of SUR labels. Then, we propose an MLP-Mixer-based network to predict SUR and SMR by leveraging and fusing the extracted multi-layer features. We introduce a Difference Feature Residual Learning (DFRL) module to learn more discriminative difference features. We further use a Multi-Head Attention Aggregation and Pooling (MHAAP) layer to aggregate difference features and reduce their redundancy. Experimental results indicate that the proposed model significantly outperforms state-of-the-art SUR and SMR prediction methods. Moreover, our joint learning scheme of human and machine perceptual quality prediction tasks is effective at improving the performance of both.

TL-Training: A Task-Feature-Based Framework for Training Large Language Models in Tool Use

Large language models (LLMs) achieve remarkable advancements by leveraging tools to interact with external environments, a critical step toward generalized AI. However, the standard supervised fine-tuning (SFT) approach, which relies on large-scale datasets, often overlooks task-specific characteristics in tool use, leading to performance bottlenecks. To address this issue, we analyze three existing LLMs and uncover key insights: training data can inadvertently impede tool-use behavior, token importance is distributed unevenly, and errors in tool calls fall into a small set of distinct categories. Building on these findings, we propose TL-Training, a task-feature-based framework that mitigates the effects of suboptimal training data, dynamically adjusts token weights to prioritize key tokens during SFT, and incorporates a robust reward mechanism tailored to error categories, optimized through proximal policy optimization. We validate TL-Training by training CodeLLaMA-2-7B and evaluating it on four diverse open-source test sets. Our results demonstrate that the LLM trained by our method matches or surpasses both open- and closed-source LLMs in tool-use performance using only 1,217 training data points. Additionally, our method enhances robustness in noisy environments and improves general task performance, offering a scalable and efficient paradigm for tool-use training in LLMs. The code and data are available at https://github.com/Junjie-Ye/TL-Training.