Jishen Zhao

Yun Joon Soh, Hanxian Huang, Yuandong Tian et al.

Supporting longer context for Large Language Models (LLM) is a promising direction to advance LLMs. As training a model for a longer context window is computationally expensive, many alternative solutions, such as Retrieval Augmented Generation (RAG), have been used. However, most existing RAG methods adopt embedding-based retrieval that falls short on long contexts. To address such challenges, we propose an attention-based retrieval technique, You Only Use Reactive Attention slice (YOURA). YOURA leverages a novel retrieval heuristic called reaction score to rank the relevance of each sentence in the input context with the query sentence. Intuitively, we measure how the per-token attention score "reacts" to the query and greedily retrieves the most reactive sentences. Internally, YOURA generates a token-indexed vector (called reaction vector) for the whole input context. To map each sentence to the token-indexed vector, we propose an Embedding-Agnostic Sentence Yield (EASY), a best-effort token wiggling algorithm. We evaluate our retrieval technique on three open-source pre-trained LLM models across six LongBench QA datasets. Our technique achieves up to 30% vLLM inference throughput improvement for serving long-context queries with a nearly identical quality score to the simple yet effective truncate-middle approach.

Haolan Liu, Liangjun Zhang, Siva Kumar Sastry Hari et al.

Generating safety-critical scenarios is essential for testing and verifying the safety of autonomous vehicles. Traditional optimization techniques suffer from the curse of dimensionality and limit the search space to fixed parameter spaces. To address these challenges, we propose a deep reinforcement learning approach that generates scenarios by sequential editing, such as adding new agents or modifying the trajectories of the existing agents. Our framework employs a reward function consisting of both risk and plausibility objectives. The plausibility objective leverages generative models, such as a variational autoencoder, to learn the likelihood of the generated parameters from the training datasets; It penalizes the generation of unlikely scenarios. Our approach overcomes the dimensionality challenge and explores a wide range of safety-critical scenarios. Our evaluation demonstrates that the proposed method generates safety-critical scenarios of higher quality compared with previous approaches.

Hui Shi, Yupeng Gu, Yitong Zhou et al.

User embeddings (vectorized representations of a user) are essential in recommendation systems. Numerous approaches have been proposed to construct a representation for the user in order to find similar items for retrieval tasks, and they have been proven effective in industrial recommendation systems as well. Recently people have discovered the power of using multiple embeddings to represent a user, with the hope that each embedding represents the user's interest in a certain topic. With multi-interest representation, it's important to model the user's preference over the different topics and how the preference change with time. However, existing approaches either fail to estimate the user's affinity to each interest or unreasonably assume every interest of every user fades with an equal rate with time, thus hurting the recall of candidate retrieval. In this paper, we propose the Multi-Interest Preference (MIP) model, an approach that not only produces multi-interest for users by using the user's sequential engagement more effectively but also automatically learns a set of weights to represent the preference over each embedding so that the candidates can be retrieved from each interest proportionally. Extensive experiments have been done on various industrial-scale datasets to demonstrate the effectiveness of our approach.

Haolan Liu, Jishen Zhao, Liangjun Zhang

Learning-based approaches to autonomous vehicle planners have the potential to scale to many complicated real-world driving scenarios by leveraging huge amounts of driver demonstrations. However, prior work only learns to estimate a single planning trajectory, while there may be multiple acceptable plans in real-world scenarios. To solve the problem, we propose an interpretable neural planner to regress a heatmap, which effectively represents multiple potential goals in the bird's-eye view of an autonomous vehicle. The planner employs an adaptive Gaussian kernel and relaxed hourglass loss to better capture the uncertainty of planning problems. We also use a negative Gaussian kernel to add supervision to the heatmap regression, enabling the model to learn collision avoidance effectively. Our systematic evaluation on the Lyft Open Dataset across a diverse range of real-world driving scenarios shows that our model achieves a safer and more flexible driving performance than prior works.

Zhongkai Yu, Yichen Lin, Chenyang Zhou et al.

Existing API-based agentic systems for RTL code generation are fundamentally misaligned with industrial practice: they assume a golden testbench is available at generation time, rely on closed-source APIs incompatible with chip vendors' air-gapped security requirements, and cannot be trained on vendors' proprietary RTL codebases, leaving valuable internal data unused. Recent self-trained models address the deployment constraint but remain single-turn generators that overlook the critical role of verification in real industrial flows. To bridge these gaps, we present ChipMATE, the first self-trained multi-agent framework for RTL generation. Inspired by industrial practice where correctness emerges from cross-comparison between independently written RTL modules and reference models, ChipMATE pairs a Verilog agent with a Python reference-model agent that mutually verify each other's outputs without any golden oracle. We design a backtrack-based inference workflow to prevent error propagation across turns, and a two-stage training pipeline that first trains each agent individually to saturate its code-generation capability, then trains the team jointly to collaborate effectively. To support the training, we further build a hybrid data-generation framework that produces 64.4K high-quality reference model training samples. ChipMATE achieves 75.0\% and 80.1\% pass@1 on VerilogEval V2 with 4B and 9B base models, outperforming all existing self-trained models and even DeepSeek V4 with 1600B parameters. Our code and model weights are publicly available in https://github.com/zhongkaiyu/ChipMATE.

Hejia Zhang, Zhongming Yu, Chia-Tung Ho et al.

Execution-aware LLM agents offer a promising paradigm for learning from tool feedback, but such feedback is often expensive and slow to obtain, making online reinforcement learning (RL) impractical. High-coverage hardware verification exemplifies this challenge due to its reliance on industrial simulators and non-differentiable execution signals. We propose LLM4Cov, an offline agent-learning framework that models verification as memoryless state transitions guided by deterministic evaluators. Building on this formulation, we introduce execution-validated data curation, policy-aware agentic data synthesis, and worst-state-prioritized sampling to enable scalable learning under execution constraints. We further curate a reality-aligned benchmark adapted from an existing verification suite through a revised evaluation protocol. Using the proposed pipeline, a compact 4B-parameter model achieves 69.2% coverage pass rate under agentic evaluation, outperforming its teacher by 5.3% and demonstrating competitive performance against models an order of magnitude larger.

Zhongkai Yu, Chenyang Zhou, Yichen Lin et al.

While Large Language Models (LLMs) show significant potential in hardware engineering, current benchmarks suffer from saturation and limited task diversity, failing to reflect LLMs' performance in real industrial workflows. To address this gap, we propose a comprehensive benchmark for AI-aided chip design that rigorously evaluates LLMs across three critical tasks: Verilog generation, debugging, and reference model generation. Our benchmark features 44 realistic modules with complex hierarchical structures, 89 systematic debugging cases, and 132 reference model samples across Python, SystemC, and CXXRTL. Evaluation results reveal substantial performance gaps, with state-of-the-art Claude-4.5-opus achieving only 30.74\% on Verilog generation and 13.33\% on Python reference model generation, demonstrating significant challenges compared to existing saturated benchmarks where SOTA models achieve over 95\% pass rates. Additionally, to help enhance LLM reference model generation, we provide an automated toolbox for high-quality training data generation, facilitating future research in this underexplored domain. Our code is available at https://github.com/zhongkaiyu/ChipBench.git.

Yujie Zhao, Boqin Yuan, Junbo Huang et al.

Large Language Models (LLMs) are deployed as autonomous agents in increasingly complex applications, where enabling long-horizon memory is critical for achieving strong performance. However, a significant gap exists between practical applications and current evaluation standards for agent memory: existing benchmarks primarily focus on dialogue-centric, human-agent interactions. In reality, agent memory consists of a continuous stream of agent-environment interactions that are primarily composed of machine-generated representations. To bridge this gap, we introduce AMA-Bench (Agent Memory with Any length), which evaluates long-horizon memory for LLMs in real agentic applications. It features two key components: (1) a set of real-world agentic trajectories across representative agentic applications, paired with expert-curated QA, and (2) a set of synthetic agentic trajectories that scale to arbitrary horizons, paired with rule-based QA. Our comprehensive study shows that existing memory systems underperform on AMA-Bench primarily because they lack causality and objective information and are constrained by the lossy nature of similarity-based retrieval employed by many memory systems. To address these limitations, we propose AMA-Agent, an effective memory system featuring a causality graph and tool-augmented retrieval. Our results demonstrate that AMA-Agent achieves 57.22% average accuracy on AMA-Bench, surpassing the strongest memory system baselines by 11.16%.

Xin Chen, Hanxian Huang, Yanjun Gao et al.

Knowledge distillation, the technique of transferring knowledge from large, complex models to smaller ones, marks a pivotal step towards efficient AI deployment. Distilling Step-by-Step~(DSS), a novel method utilizing chain-of-thought~(CoT) distillation, has demonstrated promise by imbuing smaller models with the superior reasoning capabilities of their larger counterparts. In DSS, the distilled model acquires the ability to generate rationales and predict labels concurrently through a multi-task learning framework. However, DSS overlooks the intrinsic relationship between the two training tasks, leading to ineffective integration of CoT knowledge with the task of label prediction. To this end, we investigate the mutual relationship of the two tasks from Information Bottleneck perspective and formulate it as maximizing the mutual information of the representation features of the two tasks. We propose a variational approach to solve this optimization problem using a learning-based method. Our experimental results across four datasets demonstrate that our method outperforms the state-of-the-art DSS. Our findings offer insightful guidance for future research on language model distillation as well as applications involving CoT. Codes are available at \url{https://github.com/xinchen9/cot_distillation_ACL2024}.

Zhongming Yu, Hejia Zhang, Yujie Zhao et al.

Recent developments in Large Language Model (LLM) agents are revolutionizing Autonomous Software Engineering (ASE), enabling automated coding, problem fixes, and feature improvements. However, localization -- precisely identifying software problems by navigating to relevant code sections -- remains a significant challenge. Current approaches often yield suboptimal results due to a lack of effective integration between LLM agents and precise code search mechanisms. This paper introduces OrcaLoca, an LLM agent framework that improves accuracy for software issue localization by integrating priority-based scheduling for LLM-guided action, action decomposition with relevance scoring, and distance-aware context pruning. Experimental results demonstrate that OrcaLoca becomes the new open-source state-of-the-art (SOTA) in function match rate (65.33%) on SWE-bench Lite. It also improves the final resolved rate of an open-source framework by 6.33 percentage points through its patch generation integration.

Zhongming Yu, Naicheng Yu, Hejia Zhang et al.

As LLM agents evolve into collaborative multi-agent systems, their memory requirements grow rapidly in complexity. This position paper frames multi-agent memory as a computer architecture problem. We distinguish shared and distributed memory paradigms, propose a three-layer memory hierarchy (I/O, cache, and memory), and identify two critical protocol gaps: cache sharing across agents and structured memory access control. We argue that the most pressing open challenge is multi-agent memory consistency. Our architectural framing provides a foundation for building reliable, scalable multi-agent systems.

Yujie Zhao, Hejia Zhang, Hanxian Huang et al.

The automatic generation of RTL code (e.g., Verilog) through natural language instructions has emerged as a promising direction with the advancement of large language models (LLMs). However, producing RTL code that is both syntactically and functionally correct remains a significant challenge. Existing single-LLM-agent approaches face substantial limitations because they must navigate between various programming languages and handle intricate generation, verification, and modification tasks. To address these challenges, this paper introduces MAGE, the first open-source multi-agent AI system designed for robust and accurate Verilog RTL code generation. We propose a novel high-temperature RTL candidate sampling and debugging system that effectively explores the space of code candidates and significantly improves the quality of the candidates. Furthermore, we design a novel Verilog-state checkpoint checking mechanism that enables early detection of functional errors and delivers precise feedback for targeted fixes, significantly enhancing the functional correctness of the generated RTL code. MAGE achieves a 95.7% rate of syntactic and functional correctness code generation on VerilogEval-Human 2 benchmark, surpassing the state-of-the-art Claude-3.5-sonnet by 23.3 %, demonstrating a robust and reliable approach for AI-driven RTL design workflows.

Wentao Ni, Kangqi Zhang, Zhongming Yu et al.

As long-context inference becomes central to large language models (LLMs), attention over growing key-value caches emerges as a dominant decoding bottleneck, motivating sparse attention for scalable inference. Fixed-budget top-k sparse attention cannot adapt to heterogeneous attention distributions across heads and layers, whereas top-p sparse attention directly preserves attention mass and provides stronger accuracy guarantees. Existing top-p methods, however, fail to jointly optimize top-p accuracy, selection overhead, and sparse attention cost, which limits their overall efficiency. We present Double-P, a hierarchical sparse attention framework that optimizes all three stages. Double-P first performs coarse-grained top-p estimation at the cluster level using size-weighted centroids, then adaptively refines computation through a second top-p stage that allocates token-level attention only when needed. Across long-context benchmarks, Double-P consistently achieves near-zero accuracy drop, reducing attention computation overhead by up to 1.8x and delivers up to 1.3x end-to-end decoding speedup over state-of-the-art fixed-budget sparse attention methods.

Yaolun Zhang, Yujie Zhao, Nan Wang et al.

Automatic multi-agent systems aim to instantiate agent workflows without relying on manually designed or fixed orchestration. However, existing automatic MAS approaches remain only partially adaptive: they either perform training-free test-time search or optimize the meta-level designer while keeping downstream execution agents frozen, which creating a frozen-executor ceiling and leaving the end-to-end training of self-designing and self-executing agentic models unexplored. To address this, we introduce MetaAgent-X, an end-to-end reinforcement learning framework that jointly optimizes automatic MAS design and execution. MetaAgent-X enables script-based MAS generation, execution rollout collection, and credit assignment for both designer and executor trajectories. To support stable and scalable optimization, we propose Executor Designer Hierarchical Rollout and Stagewise Co-evolution to improve training stability and expose the dynamics of designer-executor co-evolution. MetaAgent-X consistently outperforms existing automatic MAS baselines, achieving up to 21.7% gains. Comprehensive ablations show that both designer and executor improve throughout training, and that effective automatic MAS learning follows a stagewise co-evolution process. These results establish end-to-end trainable automatic MAS as a practical paradigm for building self-designing and self-executing agentic models.

Yujie Zhao, Lanxiang Hu, Yang Wang et al.

Multi-agent systems (MAS) and reinforcement learning (RL) are widely used to enhance the agentic capabilities of large language models (LLMs). MAS improves task performance through role-based orchestration, while RL uses environmental rewards to learn stronger policies, such as GRPO-style optimization. However, applying on-policy RL to MAS remains underexplored and presents unique challenges. Algorithmically, standard GRPO grouping assumptions break down because prompts vary by role and by turn. System-wise, the training stack must support MAS-workflow rollouts and on-policy updates for both single-policy and multi-policy models. We propose AT-GRPO, which includes (i) an agent- and turn-wise grouped RL algorithm tailored to MAS and (ii) a training system that supports both single- and multi-policy regimes. Across game, planning, coding, and math tasks, AT-GRPO delivers substantial gains. On long-horizon planning, it increases accuracy from a 14.0 to 47.0 percent single-agent RL baseline to 96.0 to 99.5 percent. It also improves reasoning performance, with average gains of 3.87 to 7.62 percent on coding tasks and 9.0 to 17.93 percent on math. Code and environments are available at: https://github.com/pettingllms-ai/PettingLLMs.

Yujie Zhao, Zhijing Wu, Hejia Zhang et al.

LLM-assisted hardware verification is gaining substantial attention due to its potential to significantly reduce the cost and effort of crafting effective testbenches. It also serves as a critical enabler for LLM-aided end-to-end hardware language design. However, existing current LLMs often struggle with Register Transfer Level (RTL) code generation, resulting in testbenches that exhibit functional errors in Hardware Description Languages (HDL) logic. Motivated by the strong performance of LLMs in Python code generation under inference-time sampling strategies, and their promising capabilities as judge agents, we propose PRO-V a fully program generation multi-agent system for robust RTL verification. Pro-V incorporates an efficient best-of-n iterative sampling strategy to enhance the correctness of generated testbenches. Moreover, it introduces an LLM-as-a-judge aid validation framework featuring an automated prompt generation pipeline. By converting rule-based static analysis from the compiler into natural language through in-context learning, this pipeline enables LLMs to assist the compiler in determining whether verification failures stem from errors in the RTL design or the testbench. PRO-V attains a verification accuracy of 87.17% on golden RTL implementations and 76.28% on RTL mutants. Our code is open-sourced at https://github.com/stable-lab/Pro-V.

Yun Joon Soh, Jishen Zhao

The explosion of open-sourced models and Question-Answering (QA) datasets emphasizes the importance of automated QA evaluation. We studied the statistics of the existing evaluation metrics for a better understanding of their limitations. By measuring the correlation coefficients of each evaluation metric concerning human-like evaluation score, we observed the following: (1) existing metrics have a high correlation among them concerning the question type (e.g., single word, single phrase, etc.), (2) no single metric can adequately estimate the human-like evaluation. As a potential solution, we discuss how a Mixture Of Grader could potentially improve the auto QA evaluator quality.

Hanxian Huang, Tarique Siddiqui, Rana Alotaibi et al. · microsoft-research

Database systems often rely on historical query traces to perform workload-based performance tuning. However, real production workloads are time-evolving, making historical queries ineffective for optimizing future workloads. To address this challenge, we propose SIBYL, an end-to-end machine learning-based framework that accurately forecasts a sequence of future queries, with the entire query statements, in various prediction windows. Drawing insights from real-workloads, we propose template-based featurization techniques and develop a stacked-LSTM with an encoder-decoder architecture for accurate forecasting of query workloads. We also develop techniques to improve forecasting accuracy over large prediction windows and achieve high scalability over large workloads with high variability in arrival rates of queries. Finally, we propose techniques to handle workload drifts. Our evaluation on four real workloads demonstrates that SIBYL can forecast workloads with an $87.3\%$ median F1 score, and can result in $1.7\times$ and $1.3\times$ performance improvement when applied to materialized view selection and index selection applications, respectively.

Hanxian Huang, Jishen Zhao

The increasing adoption of WebAssembly (Wasm) for performance-critical and security-sensitive tasks drives the demand for WebAssembly program comprehension and reverse engineering. Recent studies have introduced machine learning (ML)-based WebAssembly reverse engineering tools. Yet, the generalization of task-specific ML solutions remains challenging, because their effectiveness hinges on the availability of an ample supply of high-quality task-specific labeled data. Moreover, previous works overlook the high-level semantics present in source code and its documentation. Acknowledging the abundance of available source code with documentation, which can be compiled into WebAssembly, we propose to learn representations of them concurrently and harness their mutual relationships for effective WebAssembly reverse engineering. In this paper, we present WasmRev, the first multi-modal pre-trained language model for WebAssembly reverse engineering. WasmRev is pre-trained using self-supervised learning on a large-scale multi-modal corpus encompassing source code, code documentation and the compiled WebAssembly, without requiring labeled data. WasmRev incorporates three tailored multi-modal pre-training tasks to capture various characteristics of WebAssembly and cross-modal relationships. WasmRev is only trained once to produce general-purpose representations that can broadly support WebAssembly reverse engineering tasks through few-shot fine-tuning with much less labeled data, improving data efficiency. We fine-tune WasmRev onto three important reverse engineering tasks: type recovery, function purpose identification and WebAssembly summarization. Our results show that WasmRev pre-trained on the corpus of multi-modal samples establishes a robust foundation for these tasks, achieving high task accuracy and outperforming the state-of-the-art ML methods for WebAssembly reverse engineering.

Yiping Wang, Hanxian Huang, Yifang Chen et al.

While Large language models (LLMs) have advanced natural language processing tasks, their growing computational and memory demands make deployment on resource-constrained devices like mobile phones increasingly challenging. In this paper, we propose SHARP (SHaring Adjacent Layers with Recovery Parameters), a novel approach to accelerate LLM inference by sharing parameters across adjacent layers, thus reducing memory load overhead, while introducing low-rank recovery parameters to maintain performance. Inspired by observations that consecutive layers have similar outputs, SHARP employs a two-stage recovery process: Single Layer Warmup (SLW), and Supervised Fine-Tuning (SFT). The SLW stage aligns the outputs of the shared layers using L_2 loss, providing a good initialization for the following SFT stage to further restore the model performance. Extensive experiments demonstrate that SHARP can recover the model's perplexity on various in-distribution tasks using no more than 50k fine-tuning data while reducing the number of stored MLP parameters by 38% to 65%. We also conduct several ablation studies of SHARP and show that replacing layers towards the later parts of the model yields better performance retention, and that different recovery parameterizations perform similarly when parameter counts are matched. Furthermore, SHARP saves 42.8% in model storage and reduces the total inference time by 42.2% compared to the original Llama2-7b model on mobile devices. Our results highlight SHARP as an efficient solution for reducing inference costs in deploying LLMs without the need for pretraining-scale resources.

Zhongming Yu, Genghan Zhang, Hanxian Huang et al.

In recent years, Graph Neural Networks (GNNs) have ignited a surge of innovation, significantly enhancing the processing of geometric data structures such as graphs, point clouds, and meshes. As the domain continues to evolve, a series of frameworks and libraries are being developed to push GNN efficiency to new heights. While graph-centric libraries have achieved success in the past, the advent of efficient tensor compilers has highlighted the urgent need for tensor-centric libraries. Yet, efficient tensor-centric frameworks for GNNs remain scarce due to unique challenges and limitations encountered when implementing segment reduction in GNN contexts. We introduce GeoT, a cutting-edge tensor-centric library designed specifically for GNNs via efficient segment reduction. GeoT debuts innovative parallel algorithms that not only introduce new design principles but also expand the available design space. Importantly, GeoT is engineered for straightforward fusion within a computation graph, ensuring compatibility with contemporary tensor-centric machine learning frameworks and compilers. Setting a new performance benchmark, GeoT marks a considerable advancement by showcasing an average operator speedup of 1.80x and an end-to-end speedup of 1.68x.

Hui Shi, Sicun Gao, Yuandong Tian et al.

Long Short-Term Memory (LSTM) and Transformers are two popular neural architectures used for natural language processing tasks. Theoretical results show that both are Turing-complete and can represent any context-free language (CFL).In practice, it is often observed that Transformer models have better representation power than LSTM. But the reason is barely understood. We study such practical differences between LSTM and Transformer and propose an explanation based on their latent space decomposition patterns. To achieve this goal, we introduce an oracle training paradigm, which forces the decomposition of the latent representation of LSTM and the Transformer and supervises with the transitions of the Pushdown Automaton (PDA) of the corresponding CFL. With the forced decomposition, we show that the performance upper bounds of LSTM and Transformer in learning CFL are close: both of them can simulate a stack and perform stack operation along with state transitions. However, the absence of forced decomposition leads to the failure of LSTM models to capture the stack and stack operations, while having a marginal impact on the Transformer model. Lastly, we connect the experiment on the prototypical PDA to a real-world parsing task to re-verify the conclusions

Cheng Fu, Huili Chen, Haolan Liu et al.

Reverse engineering of binary executables is a critical problem in the computer security domain. On the one hand, malicious parties may recover interpretable source codes from the software products to gain commercial advantages. On the other hand, binary decompilation can be leveraged for code vulnerability analysis and malware detection. However, efficient binary decompilation is challenging. Conventional decompilers have the following major limitations: (i) they are only applicable to specific source-target language pair, hence incurs undesired development cost for new language tasks; (ii) their output high-level code cannot effectively preserve the correct functionality of the input binary; (iii) their output program does not capture the semantics of the input and the reversed program is hard to interpret. To address the above problems, we propose Coda, the first end-to-end neural-based framework for code decompilation. Coda decomposes the decompilation task into two key phases: First, Coda employs an instruction type-aware encoder and a tree decoder for generating an abstract syntax tree (AST) with attention feeding during the code sketch generation stage. Second, Coda then updates the code sketch using an iterative error correction machine guided by an ensembled neural error predictor. By finding a good approximate candidate and then fixing it towards perfect, Coda achieves superior performance compared to baseline approaches. We assess Coda's performance with extensive experiments on various benchmarks. Evaluation results show that Coda achieves an average of 82\% program recovery accuracy on unseen binary samples, where the state-of-the-art decompilers yield 0\% accuracy. Furthermore, Coda outperforms the sequence-to-sequence model with attention by a margin of 70\% program accuracy.

Hengyu Zhao, Yubo Zhang, Pingfan Meng et al.

Recently, autonomous driving development ignited competition among car makers and technical corporations. Low-level automation cars are already commercially available. But high automated vehicles where the vehicle drives by itself without human monitoring is still at infancy. Such autonomous vehicles (AVs) rely on the computing system in the car to to interpret the environment and make driving decisions. Therefore, computing system design is essential particularly in enhancing the attainment of driving safety. However, to our knowledge, no clear guideline exists so far regarding safety-aware AV computing system and architecture design. To understand the safety requirement of AV computing system, we performed a field study by running industrial Level-4 autonomous driving fleets in various locations, road conditions, and traffic patterns. The field study indicates that traditional computing system performance metrics, such as tail latency, average latency, maximum latency, and timeout, cannot fully satisfy the safety requirement for AV computing system design. To address this issue, we propose a `safety score' as a primary metric for measuring the level of safety in AV computing system design. Furthermore, we propose a perception latency model, which helps architects estimate the safety score of given architecture and system design without physically testing them in an AV. We demonstrate the use of our safety score and latency model, by developing and evaluating a safety-aware AV computing system computation hardware resource management scheme.

Cheng Fu, Shilin Zhu, Hao Su et al.

Binarized Neural Network (BNN) removes bitwidth redundancy in classical CNN by using a single bit (-1/+1) for network parameters and intermediate representations, which has greatly reduced the off-chip data transfer and storage overhead. However, a large amount of computation redundancy still exists in BNN inference. By analyzing local properties of images and the learned BNN kernel weights, we observe an average of $\sim$78% input similarity and $\sim$59% weight similarity among weight kernels, measured by our proposed metric in common network architectures. Thus there does exist redundancy that can be exploited to further reduce the amount of on-chip computations. Motivated by the observation, in this paper, we proposed two types of fast and energy-efficient architectures for BNN inference. We also provide analysis and insights to pick the better strategy of these two for different datasets and network models. By reusing the results from previous computation, much cycles for data buffer access and computations can be skipped. By experiments, we demonstrate that 80% of the computation and 40% of the buffer access can be skipped by exploiting BNN similarity. Thus, our design can achieve 17% reduction in total power consumption, 54% reduction in on-chip power consumption and 2.4$\times$ maximum speedup, compared to the baseline without applying our reuse technique. Our design also shows 1.9$\times$ more area-efficiency compared to state-of-the-art BNN inference design. We believe our deployment of BNN on FPGA leads to a promising future of running deep learning models on mobile devices.