Hasil untuk "Speculative philosophy"

Yida Zhang, Zhiyong Gao, Shuaibing Yue et al.

Recent advancements and widespread adoption of Large Language Models (LLMs) in both industry and academia have catalyzed significant demand for LLM serving. However, traditional cloud services incur high costs, while on-device inference alone faces challenges due to limited resources. Edge-cloud collaboration emerges as a key research direction to combine the strengths of both paradigms, yet efficiently utilizing limited network bandwidth while fully leveraging and balancing the computational capabilities of edge devices and the cloud remains an open problem. To address these challenges, we propose Pipelined Collaborative Speculative Decoding Framework (PicoSpec), a novel, general-purpose, and training-free speculative decoding framework for LLM edge-cloud collaborative inference. We design an asynchronous pipeline that resolves the mutual waiting problem inherent in vanilla speculative decoding within edge collaboration scenarios, which concurrently executes a Small Language Model (SLM) on the edge device and a LLM in the cloud. Meanwhile, to mitigate the significant communication latency caused by transmitting vocabulary distributions, we introduce separate rejection sampling with sparse compression, which completes the rejection sampling with only a one-time cost of transmitting the compressed vocabulary. Experimental results demonstrate that our solution outperforms baseline and existing methods, achieving up to 2.9 speedup.

Rahul Thomas, Arka Pal

The goal of $L$-step speculative decoding is to accelerate autoregressive decoding of a target model by using a cheaper draft model to generate a candidate path of $L$ tokens. Based on a verification algorithm involving target and draft model probabilities, a prefix of the candidate sequence is accepted, and an additional correction token is sampled from a residual distribution to ensure that the final output adheres to the target distribution. While standard speculative decoding uses a verification algorithm which is independent at each token on the path, a recent extension called block verification uses a joint condition involving all sampled on-path probabilities. Block verification (BV) was shown to be optimal over all verification algorithms which use only on-path probabilities, improving on standard speculative decoding. In this work, we first show that block verification is optimal even over verification algorithms that use off-path probabilities, by constructing an information-agnostic linear program (LP). Further, we can extend our LP to the setting where the draft model samples multiple candidate paths, and use it to construct a natural class of multi-path block verification generalizations. While computing the optimal algorithm in this class is not tractable, by considering a stricter class of greedy algorithms, we can formulate an efficient method called greedy multi-path block verification (GBV). Empirically, GBV can improve block efficiency by over 30% and reduce decoding walltimes by over 15% relative to BV. On Llama-3 70B, GBV can improve the end-to-end decoding throughput over SOTA multi-path verification methods by more than 15%.

Shubhangi Upasani, Ravi Shanker Raju, Bo Li et al.

Prompt length is a major bottleneck in agentic large language model (LLM) workloads, where repeated inference steps and multi-call loops incur substantial prefill cost. Recent work on speculative prefill demonstrates that attention-based token importance estimation can enable training-free prompt compression, but this assumes the existence of a draft model that shares the same tokenizer as the target model. In practice, however, agentic pipelines frequently employ models without any smaller in-family draft model. In this work, we study cross-family speculative prefill, where a lightweight draft model from one model family is used to perform prompt compression for a target model from a different family. Using the same speculative prefill mechanism as prior work, we evaluate a range of cross-family draft-target combinations, including Qwen, LLaMA, and DeepSeek models. Across a broad diversity of tasks, we find that attention-based token importance estimation transfers reliably across different model families despite differences in model architectures and tokenizers between draft and target models. Cross-model prompt compression largely retains 90~100% of full-prompt baseline performance and, in some cases, slightly improves accuracy due to denoising effects, while delivering substantial reductions in time to first token (TTFT). These results suggest that speculative prefill depends mainly on task priors and semantic structure, thus serving as a generalizable prompt compression primitive. We discuss the implications of our findings for agentic systems, where repeated long-context inference and heterogeneous model stacks make cross-model prompt compression both necessary and practical.

Siyuan He, Zhantong Zhu, Yandong He et al.

LLM inference on mobile devices faces extraneous challenges due to limited memory bandwidth and computational resources. To address these issues, speculative inference and processing-in-memory (PIM) techniques have been explored at the algorithmic and hardware levels. However, speculative inference results in more compute-intensive GEMM operations, creating new design trade-offs for existing GEMV-accelerated PIM architectures. Furthermore, there exists a significant amount of redundant draft tokens in tree-based speculative inference, necessitating efficient token management schemes to minimize energy consumption. In this work, we present LP-Spec, an architecture-dataflow co-design leveraging hybrid LPDDR5 performance-enhanced PIM architecture with draft token pruning and dynamic workload scheduling to accelerate LLM speculative inference. A near-data memory controller is proposed to enable data reallocation between DRAM and PIM banks. Furthermore, a data allocation unit based on the hardware-aware draft token pruner is developed to minimize energy consumption and fully exploit parallel execution opportunities. Compared to end-to-end LLM inference on other mobile solutions such as mobile NPUs or GEMV-accelerated PIMs, our LP-Spec achieves 13.21x, 7.56x, and 99.87x improvements in performance, energy efficiency, and energy-delay-product (EDP). Compared with prior AttAcc PIM and RTX 3090 GPU, LP-Spec can obtain 12.83x and 415.31x EDP reduction benefits.

Ziyi Zhang, Ziheng Jiang, Chengquan Jiang et al.

Low-latency decoding for large language models (LLMs) is crucial for applications like chatbots and code assistants, yet generating long outputs remains slow in single-query settings. Prior work on speculative decoding (which combines a small draft model with a larger target model) and tensor parallelism has each accelerated decoding. However, conventional approaches fail to apply both simultaneously due to imbalanced compute requirements (between draft and target models), KV-cache inconsistencies, and communication overheads under small-batch tensor-parallelism. This paper introduces SwiftSpec, a system that targets ultra-low latency for LLM decoding. SwiftSpec redesigns the speculative decoding pipeline in an asynchronous and disaggregated manner, so that each component can be scaled flexibly and remove draft overhead from the critical path. To realize this design, SwiftSpec proposes parallel tree generation, tree-aware KV cache management, and fused, latency-optimized kernels to overcome the challenges listed above. Across 5 model families and 6 datasets, SwiftSpec achieves an average of 1.75x speedup over state-of-the-art speculative decoding systems and, as a highlight, serves Llama3-70B at 348 tokens/s on 8 Nvidia Hopper GPUs, making it the fastest known system for low-latency LLM serving at this scale.

Jialiang Kang, Han Shu, Wenshuo Li et al.

Speculative decoding is a widely adopted technique for accelerating inference in large language models (LLMs), yet its application to vision-language models (VLMs) remains underexplored, with existing methods achieving only modest speedups (<1.5x). This gap is increasingly significant as multimodal capabilities become central to large-scale models. We hypothesize that large VLMs can effectively filter redundant image information layer by layer without compromising textual comprehension, whereas smaller draft models struggle to do so. To address this, we introduce Vision-Aware Speculative Decoding (ViSpec), a novel framework tailored for VLMs. ViSpec employs a lightweight vision adaptor module to compress image tokens into a compact representation, which is seamlessly integrated into the draft model's attention mechanism while preserving original image positional information. Additionally, we extract a global feature vector for each input image and augment all subsequent text tokens with this feature to enhance multimodal coherence. To overcome the scarcity of multimodal datasets with long assistant responses, we curate a specialized training dataset by repurposing existing datasets and generating extended outputs using the target VLM with modified prompts. Our training strategy mitigates the risk of the draft model exploiting direct access to the target model's hidden states, which could otherwise lead to shortcut learning when training solely on target model outputs. Extensive experiments validate ViSpec, achieving, to our knowledge, the first substantial speedup in VLM speculative decoding. Code is available at https://github.com/KangJialiang/ViSpec.

Baohao Liao, Yuhui Xu, Hanze Dong et al.

We introduce Reward-Guided Speculative Decoding (RSD), a novel framework aimed at improving the efficiency of inference in large language models (LLMs). RSD synergistically combines a lightweight draft model with a more powerful target model, incorporating a controlled bias to prioritize high-reward outputs, in contrast to existing speculative decoding methods that enforce strict unbiasedness. RSD employs a process reward model to evaluate intermediate decoding steps and dynamically decide whether to invoke the target model, optimizing the trade-off between computational cost and output quality. We theoretically demonstrate that a threshold-based mixture strategy achieves an optimal balance between resource utilization and performance. Extensive evaluations on challenging reasoning benchmarks, including Olympiad-level tasks, show that RSD delivers significant efficiency gains against decoding with the target model only (up to 4.4x fewer FLOPs), while achieving significant better accuracy than parallel decoding method on average (up to +3.5). These results highlight RSD as a robust and cost-effective approach for deploying LLMs in resource-intensive scenarios. The code is available at https://github.com/BaohaoLiao/RSD.

Shukai Gong, Yiyang Fu, Fengyuan Ran et al.

We propose TPP-SD, a novel approach that accelerates Transformer temporal point process (TPP) sampling by adapting speculative decoding (SD) techniques from language models. By identifying the structural similarities between thinning algorithms for TPPs and speculative decoding for language models, we develop an efficient sampling framework that leverages a smaller draft model to generate multiple candidate events, which are then verified by the larger target model in parallel. TPP-SD maintains the same output distribution as autoregressive sampling while achieving significant acceleration. Experiments on both synthetic and real datasets demonstrate that our approach produces samples from identical distributions as standard methods, but with 2-6$\times$ speedup. Our ablation studies analyze the impact of hyperparameters such as draft length and draft model size on sampling efficiency. TPP-SD bridges the gap between powerful Transformer TPP models and the practical need for rapid sequence sampling.

Somnath Roy, Padharthi Sreekar, Srivatsa Narasimha et al.

The current study introduces a novel adaptation of speculative decoding, repurposed from generation to classification tasks. We propose a multi-model framework employing up to three lightweight worker models and a single, more robust judge model analogous to draft models and target model, respectively, in speculative decoding. The worker models, tasked with the bulk of the computation, independently predict discrete class labels for a given input. When majority worker models agree on a label, it is accepted as the final label, optimizing efficiency by bypassing the computationally expensive judge model. In cases of disagreement, the judge model intervenes to resolve the label. This approach minimizes redundant computation, leverages the redundancy of multiple workers for confidence, and confines the judge model's role to challenging cases, offering a practical balance of efficiency and accuracy. Our analysis suggests that smaller out of the box instruction/chat finetuned worker models with 3 billion parameters (hereafter, 3B) demonstrate a level of alignment with judge models comparable to that of larger finetuned worker models with 7 billion parameters (hereafter, 7B) across both simple and higher order reasoning tasks. The top performing 3B worker model pair achieve an agreement rate of approximately 80-83% for sentiment and around 50-80% for similar ticket when compared to judge models. Additionally, 3B worker models provide a speedup ranging from 2.8x to 9x relative to the judge models, while 7B worker model combinations achieve a speedup ranging from 1.28x to 0.28x

Dong Liu, Yanxuan Yu

Large Language Models (LLMs) have revolutionized natural language processing tasks, but their deployment in datacenter environments faces significant challenges due to the massive memory requirements of key-value (KV) caches. During the autoregressive decoding process, KV caches consume substantial GPU memory, limiting batch sizes and overall system throughput. To address these challenges, we propose \textbf{CXL-SpecKV}, a novel disaggregated KV-cache architecture that leverages Compute Express Link (CXL) interconnects and FPGA accelerators to enable efficient speculative execution and memory disaggregation. Our approach introduces three key innovations: (i) a CXL-based memory disaggregation framework that offloads KV-caches to remote FPGA memory with low latency, (ii) a speculative KV-cache prefetching mechanism that predicts and preloads future tokens' cache entries, and (iii) an FPGA-accelerated KV-cache compression and decompression engine that reduces memory bandwidth requirements by up to 4$\times$. When evaluated on state-of-the-art LLM models, CXL-SpecKV achieves up to 3.2$\times$ higher throughput compared to GPU-only baselines, while reducing memory costs by 2.8$\times$ and maintaining accuracy. Our system demonstrates that intelligent memory disaggregation combined with speculative execution can effectively address the memory wall challenge in large-scale LLM serving. Our code implementation has been open-sourced at https://github.com/FastLM/CXL-SpecKV.

Amund Bergland Kvalsvik, Magnus Själander

Secure speculation schemes have shown great promise in the war against speculative side-channel attacks, and will be a key building block for developing secure, high-performance architectures moving forward. As the field matures, the need for rigorous microarchitectures, and corresponding performance and cost analysis, become critical for evaluating secure schemes and for enabling their future adoption. In ShadowBinding, we present effective microarchitectures for two state-of-the-art secure schemes, uncovering and mitigating fundamental microarchitectural limitations within the analyzed schemes, and provide important design characteristics. We uncover that Speculative Taint Tracking's (STT's) rename-based taint computation must be completed in a single cycle, creating an expensive dependency chain which greatly limits performance for wider processor cores. We also introduce a novel michroarchitectural approach for STT, named STT-Issue, which, by delaying the taint computation to the issue stage, eliminates the dependency chain, achieving better instructions per cycle (IPC), timing, area, and performance results. Through a comprehensive evaluation of our STT and Non-Speculative Data Access (NDA) microarchitectural designs on the RISC-V Berkeley Out-of-Order Machine, we find that the IPC impact of in-core secure schemes is higher than previously estimated, close to 20% for the highest performance core. With insights into timing from our RTL evaluation, the performance loss, created by the combined impact of IPC and timing, becomes even greater, at 35%, 27%, and 22% for STT-Rename, STT-Issue, and NDA, respectively. If these trends were to hold for leading processor core designs, the performance impact would be well over 30%, even for the best-performing scheme.

Yingpeng Du, Tianjun Wei, Zhu Sun et al.

Large Language Models (LLMs) have been widely adopted in ranking systems such as information retrieval (IR) systems and recommender systems (RSs). To alleviate the latency of auto-regressive decoding, some studies explore the single (first) token decoding for ranking approximation, but they suffer from severe degradation in tail positions. Although speculative decoding (SD) methods can be a remedy with verification at different positions, they face challenges in ranking systems due to their left-to-right decoding paradigm. Firstly, ranking systems require strict latency constraints, but verification rounds in SD methods remain agnostic; Secondly, SD methods usually discard listwise ranking knowledge about unaccepted items in previous rounds, hindering future multi-token prediction, especially when candidate tokens are the unaccepted items. In this paper, we propose a Reinforcement Speculative Decoding method for fast ranking inference of LLMs. To meet the ranking systems' latency requirement, we propose an up-to-down decoding paradigm that employs an agent to iteratively modify the ranking sequence under a constrained budget. Specifically, we design a ranking-tailored policy optimization, actively exploring optimal multi-round ranking modification policy verified by LLMs via reinforcement learning (RL). To better approximate the target LLM under the constrained budget, we trigger the agent fully utilizing the listwise ranking knowledge about all items verified by LLMs across different rounds in RL, enhancing the modification policy of the agent. More importantly, we demonstrate the theoretical robustness and advantages of our paradigm and implementation. Experiments on both IR and RS tasks show the effectiveness of our proposed method.

Сергій Володимирович САВЧЕНКО, Олександр Олександрович МИХАЙЛЮК

This article examines the impact of Edward Said’s concept of Orientalism on the contemporary geopolitical landscape. The authors argue that while the critique of Orientalism has significantly influenced academic discourse and domestic policies in Western countries, particularly the United States, it has not led to fundamental changes in global geopolitical strategy. The paper focuses on two key aspects: the natural connection between knowledge and power, which cannot be altered by anti-colonial declarations, and the shift of colonial discourse to new regions, specifically non-Western Europe. The authors posit that the strength of a modern state lies in its ability to dynamically bifurcate its image into internal and external faces, presenting different aspects to its society and foreign partners. Using the United States as a case study, the article demonstrates how internal anti-colonial discourse coexists with an unchanging foreign policy strategy of global dominance. The authors criticize the illusions about the possibility of changing the fundamental principles of US foreign policy, which has remained consistent since 1941. The paper emphasizes the continuation of a strategy aimed at fostering global dependence on the West, despite anti-colonial rhetoric. The article explores the complex interplay between domestic movements for social justice and decolonization, and the pragmatic realities of international relations. It argues that the deconstruction of Orientalism, while challenging the legitimacy of historical colonial practices, has not transformed the core geopolitical interests of Western states. The authors contend that the apparent dichotomy between internal anti-colonial discourse and external policy reflects both the complexity and strength of states like the US, where domestic and foreign policies can be shaped by different institutional and ideological forces. The paper also discusses the expansion of the concept of liberal democracy as an export product to regions deemed lacking in popular sovereignty. In conclusion, the article suggests that the critique of Orientalism, despite its academic and cultural impact, has been effectively neutralized in geopolitical practice. This is achieved through a shift in focus to regions not traditionally considered part of the «old» Western colonies, allowing for the implementation of economic partnership policies without the moral self-flagellation associated with charges of Orientalism. The paper ultimately argues that renaming reality does not change its underlying structure, and that the postmodern belief in the power of language to transform geopolitical realities is akin to magical thinking. It concludes that the critique of Orientalism has not fundamentally altered the eternal principles and interests guiding global geopolitical strategies.

Wonseok Jeon, Mukul Gagrani, Raghavv Goel et al.

Speculative decoding is an inference-acceleration method for large language models (LLMs) where a small language model generates a draft-token sequence which is further verified by the target LLM in parallel. Recent works have advanced this method by establishing a draft-token tree, achieving superior performance over a single-sequence speculative decoding. However, those works independently generate tokens at each level of the tree, not leveraging the tree's entire diversifiability. Besides, their empirical superiority has been shown for fixed length of sequences, implicitly granting more computational resource to LLM for the tree-based methods. None of the existing works has conducted empirical studies with fixed target computational budgets despite its importance to resource-bounded devices. We present Recursive Speculative Decoding (RSD), a novel tree-based method that samples draft tokens without replacement and maximizes the diversity of the tree. During RSD's drafting, the tree is built by either Gumbel-Top-$k$ trick that draws tokens without replacement in parallel or Stochastic Beam Search that samples sequences without replacement while early-truncating unlikely draft sequences and reducing the computational cost of LLM. We empirically evaluate RSD with Llama 2 and OPT models, showing that RSD outperforms the baseline methods, consistently for fixed draft sequence length and in most cases for fixed computational budgets at LLM.

Riaz Ud Din, Salman Ahmed, Saddam Hussain Khan

Forecasting speculative stock prices is essential for effective investment risk management that drives the need for the development of innovative algorithms. However, the speculative nature, volatility, and complex sequential dependencies within financial markets present inherent challenges which necessitate advanced techniques. This paper proposes a novel framework, CAB-XDE (customized attention BiLSTM-XGB decision ensemble), for predicting the daily closing price of speculative stock Bitcoin-USD (BTC-USD). CAB-XDE framework integrates a customized bi-directional long short-term memory (BiLSTM) with the attention mechanism and the XGBoost algorithm. The customized BiLSTM leverages its learning capabilities to capture the complex sequential dependencies and speculative market trends. Additionally, the new attention mechanism dynamically assigns weights to influential features, thereby enhancing interpretability, and optimizing effective cost measures and volatility forecasting. Moreover, XGBoost handles nonlinear relationships and contributes to the proposed CAB-XDE framework robustness. Additionally, the weight determination theory-error reciprocal method further refines predictions. This refinement is achieved by iteratively adjusting model weights. It is based on discrepancies between theoretical expectations and actual errors in individual customized attention BiLSTM and XGBoost models to enhance performance. Finally, the predictions from both XGBoost and customized attention BiLSTM models are concatenated to achieve diverse prediction space and are provided to the ensemble classifier to enhance the generalization capabilities of CAB-XDE. The proposed CAB-XDE framework is empirically validated on volatile Bitcoin market, sourced from Yahoo Finance and outperforms state-of-the-art models with a MAPE of 0.0037, MAE of 84.40, and RMSE of 106.14.

Yuxuan Hu, Ke Wang, Xiaokang Zhang et al.

Speculative decoding (SD) has been demonstrated as an effective technique for lossless LLM inference acceleration. Retrieval-based SD methods, one kind of model-free method, have yielded promising speedup, but they often rely on incomplete retrieval resources, inefficient retrieval methods, and are constrained to certain domains. This paper presents a novel retrieval-based speculative decoding method that adapts suffix automaton (SAM) for efficient and accurate draft generation by utilizing common text corpus and dynamic text sequence. Unlike existing $n$-gram matching methods, SAM-Decoding finds the exact longest suffix match, achieving an average time complexity of O(1) per generation step of SAM update and suffix retrieval. It can also integrate with existing methods, adaptively selecting a draft generation strategy based on match length to generalize to broader domains. Extensive experiments on Spec-Bench show that our method is $18\%+$ faster than other retrieval-based SD methods. Additionally, when combined with advanced EAGLE-2, it provides an additional speedup of $3.28\%$ -- $11.13\%$ across various-sized LLM backbones. Our code is available at our \href{https://github.com/hyx1999/SAM-Decoding}{repository}.

Юлія Юріївна БРОДЕЦЬКА

Дослідження зосереджено на розгляді кайнонічної природи української філософії, що розкривається в потенціалі кордоцентризму. Як напрямок філософської традиції, кордоцентризм постулює проблему гносеологічних ресурсів філософського знання. Найбільш повно напрямок розкривається в філософії Г.Сковороди, де християнська традиція поєднується з українською самобутністю. Саме ця синтезність і закладає перспективи вирішення проблеми методологічної різноплановості сучасної філософії через актуалізацію фундаментальних принципів філософського пізнання – цілісності та комплексності. Таким чином, в контексті методологічної проблематики сучасної науки звернення до кардоцентризму, й безпосередньо філософії серця Г.Сковороди, є перспективним напрямком розвитку сучасної філософської традиції.

Liliane Dutra Brignol, Sabrina Cáceres

This article investigates Inumeráveis, a digital memorial that seeks to honor the victims of Covid-19 in Brazil, in order to know how the mediations of technicity and narratives, constitutive of the project, build collective memories and experiences of mourning in relation to the pandemic.The article starts with a transdisciplinary discussion about the concepts of grief and memory, crossed by the communicational perspective and digitalization. In the methodology, we combine two perspectives: media archeology and critical narrative analysis. We present part of the excavation and expedition work to know the project's production logic, its creators, the aesthetic and symbolic aspects of digital platforms, and an interpretative synthesis of the analysis of 26 texts tributes of the memorial. As a result, we highlight how the memorial provides different manifestations of the grief work and the construction of a national and political memory about the pandemic.

Eugene Y. S. Chua, Craig Callender

Programs in quantum gravity often claim that time emerges from fundamentally timeless physics. In the semiclassical time program time arises only after approximations are taken. Here we ask what justifies taking these approximations and show that time seems to sneak in when answering this question. This raises the worry that the approach is either unjustified or circular in deriving time from no-time.

Allison Wolf

With few exceptions, philosophers working on immigration have not taken up the topic of epistemic injustice, primarily, I imagine, because immigration justice is often too narrowly conceived of as encompassing moral and political concerns rather than epistemic ones. But the more I think about the injustices immigrants endure on a daily basis, the more I take this to be a mistake; epistemic injustices must be seen as a central aspect of immigration injustice too. In what follows, I will demonstrate how this is the case. More specifically, after providing an overview of the nature of epistemic injustice, I will highlight some examples of it in the lives of displaced Venezuelan immigrants in Colombia. In doing so, I hope to show why discussions about immigration injustice must include identifying and confronting epistemic wrongs.