The Bitter Lesson Learned from 2,000+ Multilingual Benchmarks

As large language models (LLMs) continue to advance in linguistic capabilities, robust multilingual evaluation has become essential for promoting equitable technological progress. This position paper examines over 2,000 multilingual (non-English) benchmarks from 148 countries, published between 2021 and 2024, to evaluate past, present, and future practices in multilingual benchmarking. Our findings reveal that, despite significant investments amounting to tens of millions of dollars, English remains significantly overrepresented in these benchmarks. Additionally, most benchmarks rely on original language content rather than translations, with the majority sourced from high-resource countries such as China, India, Germany, the UK, and the USA. Furthermore, a comparison of benchmark performance with human judgments highlights notable disparities. STEM-related tasks exhibit strong correlations with human evaluations (0.70 to 0.85), while traditional NLP tasks like question answering (e.g., XQuAD) show much weaker correlations (0.11 to 0.30). Moreover, translating English benchmarks into other languages proves insufficient, as localized benchmarks demonstrate significantly higher alignment with local human judgments (0.68) than their translated counterparts (0.47). This underscores the importance of creating culturally and linguistically tailored benchmarks rather than relying solely on translations. Through this comprehensive analysis, we highlight six key limitations in current multilingual evaluation practices, propose the guiding principles accordingly for effective multilingual benchmarking, and outline five critical research directions to drive progress in the field. Finally, we call for a global collaborative effort to develop human-aligned benchmarks that prioritize real-world applications.

LMM4LMM: Benchmarking and Evaluating Large-multimodal Image Generation with LMMs

Recent breakthroughs in large multimodal models (LMMs) have significantly advanced both text-to-image (T2I) generation and image-to-text (I2T) interpretation. However, many generated images still suffer from issues related to perceptual quality and text-image alignment. Given the high cost and inefficiency of manual evaluation, an automatic metric that aligns with human preferences is desirable. To this end, we present EvalMi-50K, a comprehensive dataset and benchmark for evaluating large-multimodal image generation, which features (i) comprehensive tasks, encompassing 2,100 extensive prompts across 20 fine-grained task dimensions, and (ii) large-scale human-preference annotations, including 100K mean-opinion scores (MOSs) and 50K question-answering (QA) pairs annotated on 50,400 images generated from 24 T2I models. Based on EvalMi-50K, we propose LMM4LMM, an LMM-based metric for evaluating large multimodal T2I generation from multiple dimensions including perception, text-image correspondence, and task-specific accuracy. Extensive experimental results show that LMM4LMM achieves state-of-the-art performance on EvalMi-50K, and exhibits strong generalization ability on other AI-generated image evaluation benchmark datasets, manifesting the generality of both the EvalMi-50K dataset and LMM4LMM metric. Both EvalMi-50K and LMM4LMM will be released at https://github.com/IntMeGroup/LMM4LMM.

TAU: Modeling Temporal Consistency Through Temporal Attentive U-Net for PPG Peak Detection

Photoplethysmography (PPG) sensors have been widely used in consumer wearable devices to monitor heart rates (HR) and heart rate variability (HRV). Despite the prevalence, PPG signals can be contaminated by motion artifacts induced from daily activities. Existing approaches mainly use the amplitude information to perform PPG peak detection. However, these approaches cannot accurately identify peaks, since motion artifacts may bring random and significant amplitude variations. To improve the performance of PPG peak detection, the time information can be used. Specifically, heart rates exhibit temporal consistency that consecutive heartbeat intervals in a normal person can have limited variations. To leverage the temporal consistency, we propose the Temporal Attentive U-Net, i.e., TAU, to accurately detect peaks from PPG signals. In TAU, we design a time module that encodes temporal consistency in temporal embeddings. We integrate the amplitude information with temporal embeddings using the attention mechanism to estimate peak labels. Our experimental results show that TAU outperforms eleven baselines on heart rate estimation by more than 22.4%. Our TAU model achieves the best performance across various Signal-to-Noise Ratio (SNR) levels. Moreover, we achieve Pearson correlation coefficients higher than 0.9 (p < 0.01) on estimating HRV features from low-noise-level PPG signals.

CoinRobot: Generalized End-to-end Robotic Learning for Physical Intelligence

Physical intelligence holds immense promise for advancing embodied intelligence, enabling robots to acquire complex behaviors from demonstrations. However, achieving generalization and transfer across diverse robotic platforms and environments requires careful design of model architectures, training strategies, and data diversity. Meanwhile existing systems often struggle with scalability, adaptability to heterogeneous hardware, and objective evaluation in real-world settings. We present a generalized end-to-end robotic learning framework designed to bridge this gap. Our framework introduces a unified architecture that supports cross-platform adaptability, enabling seamless deployment across industrial-grade robots, collaborative arms, and novel embodiments without task-specific modifications. By integrating multi-task learning with streamlined network designs, it achieves more robust performance than conventional approaches, while maintaining compatibility with varying sensor configurations and action spaces. We validate our framework through extensive experiments on seven manipulation tasks. Notably, Diffusion-based models trained in our framework demonstrated superior performance and generalizability compared to the LeRobot framework, achieving performance improvements across diverse robotic platforms and environmental conditions.

Q-Filters: Leveraging QK Geometry for Efficient KV Cache Compression

Autoregressive language models rely on a Key-Value (KV) Cache, which avoids re-computing past hidden states during generation, making it faster. As model sizes and context lengths grow, the KV Cache becomes a significant memory bottleneck, which calls for compression methods that limit its size during generation. In this paper, we discover surprising properties of Query (Q) and Key (K) vectors that allow us to efficiently approximate attention scores without computing the attention maps. We propose Q-Filters, a training-free KV Cache compression method that filters out less crucial Key-Value pairs based on a single context-agnostic projection. Contrarily to many alternatives, Q-Filters is compatible with FlashAttention, as it does not require direct access to attention weights. Experimental results in long-context settings demonstrate that Q-Filters is competitive with attention-based compression methods such as SnapKV in retrieval tasks while consistently outperforming efficient compression schemes such as Streaming-LLM in generation setups. Notably, Q-Filters achieves a 99% accuracy in the needle-in-a-haystack task with a x32 compression level while reducing the generation perplexity drop by up to 65% in text generation compared to Streaming-LLM.

Towards Widening The Distillation Bottleneck for Reasoning Models

Large Reasoning Models(LRMs) such as OpenAI o1 and DeepSeek-R1 have shown remarkable reasoning capabilities by scaling test-time compute and generating long Chain-of-Thought(CoT). Distillation--post-training on LRMs-generated data--is a straightforward yet effective method to enhance the reasoning abilities of smaller models, but faces a critical bottleneck: we found that distilled long CoT data poses learning difficulty for small models and leads to the inheritance of biases (i.e. over-thinking) when using Supervised Fine-tuning(SFT) and Reinforcement Learning(RL) methods. To alleviate this bottleneck, we propose constructing tree-based CoT data from scratch via Monte Carlo Tree Search(MCTS). We then exploit a set of CoT-aware approaches, including Thoughts Length Balance, Fine-grained DPO, and Joint Post-training Objective, to enhance SFT and RL on the construted data.

CNsum:Automatic Summarization for Chinese News Text

Obtaining valuable information from massive data efficiently has become our research goal in the era of Big Data. Text summarization technology has been continuously developed to meet this demand. Recent work has also shown that transformer-based pre-trained language models have achieved great success on various tasks in Natural Language Processing (NLP). Aiming at the problem of Chinese news text summary generation and the application of Transformer structure on Chinese, this paper proposes a Chinese news text summarization model (CNsum) based on Transformer structure, and tests it on Chinese datasets such as THUCNews. The results of the conducted experiments show that CNsum achieves better ROUGE score than the baseline models, which verifies the outperformance of the model.

Utilizing Effective Dynamic Graph Learning to Shield Financial Stability from Risk Propagation

Financial risks can propagate across both tightly coupled temporal and spatial dimensions, posing significant threats to financial stability. Moreover, risks embedded in unlabeled data are often difficult to detect. To address these challenges, we introduce GraphShield, a novel approach with three key innovations: Enhanced Cross-Domain Infor mation Learning: We propose a dynamic graph learning module to improve information learning across temporal and spatial domains. Advanced Risk Recognition: By leveraging the clustering characteristics of risks, we construct a risk recognizing module to enhance the identification of hidden threats. Risk Propagation Visualization: We provide a visualization tool for quantifying and validating nodes that trigger widespread cascading risks. Extensive experiments on two real-world and two open-source datasets demonstrate the robust performance of our framework. Our approach represents a significant advancement in leveraging artificial intelligence to enhance financial stability, offering a powerful solution to mitigate the spread of risks within financial networks.

MDEval: Evaluating and Enhancing Markdown Awareness in Large Language Models

Large language models (LLMs) are expected to offer structured Markdown responses for the sake of readability in web chatbots (e.g., ChatGPT). Although there are a myriad of metrics to evaluate LLMs, they fail to evaluate the readability from the view of output content structure. To this end, we focus on an overlooked yet important metric -- Markdown Awareness, which directly impacts the readability and structure of the content generated by these language models. In this paper, we introduce MDEval, a comprehensive benchmark to assess Markdown Awareness for LLMs, by constructing a dataset with 20K instances covering 10 subjects in English and Chinese. Unlike traditional model-based evaluations, MDEval provides excellent interpretability by combining model-based generation tasks and statistical methods. Our results demonstrate that MDEval achieves a Spearman correlation of 0.791 and an accuracy of 84.1% with human, outperforming existing methods by a large margin. Extensive experimental results also show that through fine-tuning over our proposed dataset, less performant open-source models are able to achieve comparable performance to GPT-4o in terms of Markdown Awareness. To ensure reproducibility and transparency, MDEval is open sourced at https://github.com/SWUFE-DB-Group/MDEval-Benchmark.

Application of Vision-Language Model to Pedestrians Behavior and Scene Understanding in Autonomous Driving

Autonomous driving (AD) has experienced significant improvements in recent years and achieved promising 3D detection, classification, and localization results. However, many challenges remain, e.g. semantic understanding of pedestrians' behaviors, and downstream handling for pedestrian interactions. Recent studies in applications of Large Language Models (LLM) and Vision-Language Models (VLM) have achieved promising results in scene understanding and high-level maneuver planning in diverse traffic scenarios. However, deploying the billion-parameter LLMs to vehicles requires significant computation and memory resources. In this paper, we analyzed effective knowledge distillation of semantic labels to smaller Vision networks, which can be used for the semantic representation of complex scenes for downstream decision-making for planning and control.