LLMs Can Easily Learn to Reason from Demonstrations Structure, not content, is what matters!

ArXiv.org · 2025-02-11

Large reasoning models (LRMs) tackle complex reasoning problems by following long chain-of-thoughts (Long CoT) that incorporate reflection, backtracking, and self-validation. However, the training techniques and data requirements to elicit Long CoT remain poorly understood. In this work, we find that a Large Language model (LLM) can effectively learn Long CoT reasoning through data-efficient supervised fine-tuning (SFT) and parameter-efficient low-rank adaptation (LoRA). With just 17k long CoT training samples, the Qwen2.5-32B-Instruct model achieves significant improvements on a wide range of math and coding benchmarks, including 56.7% (+40.0%) on AIME 2024 and 57.0% (+8.1%) on LiveCodeBench, competitive to the proprietary o1-preview model's score of 44.6% and 59.1%. More importantly, we find that the structure of Long CoT is critical to the learning process, whereas the content of individual reasoning steps has minimal impact. Perturbations affecting content, such as training on incorrect samples or removing reasoning keywords, have little impact on performance. In contrast, structural modifications that disrupt logical consistency in the Long CoT, such as shuffling or deleting reasoning steps, significantly degrade accuracy. For example, a model trained on Long CoT samples with incorrect answers still achieves only 3.2% lower accuracy compared to training with fully correct samples. These insights deepen our understanding of how to elicit reasoning capabilities in LLMs and highlight key considerations for efficiently training the next generation of reasoning models. This is the academic paper of our previous released Sky-T1-32B-Preview model. Codes are available at https://github.com/NovaSky-AI/SkyThought.

Rethinking the Cost of Distributed Caches for Datacenter Services

2025-11-17

This paper systematically studies the cost impact of distributed in-memory caches on datacenter services. While memory used for these caches is often perceived to be expensive, we find that the resulting CPU savings from these in-memory caches far outweigh the cost of added memory. In fact, across a variety of both synthetic and production workloads, we find that adding distributed in-memory caches can lower total operating costs by 3 – 4×, even without considering their latency benefits. These cost savings can vary significantly across various architectures, such as storage layer caches, remote lookaside caches, and in-memory linked caches. We additionally evaluate cost for two emerging scenarios: caching rich application objects and strongly consistent cache. For the former, we find that caching application objects provides outsized benefits compared to their denormalized, key-value-style variants, up to 8× compared to reading from storage. For the latter, we observe that even a minimal version check for consistency can eliminate most of the cost benefits, calling for new designs for cost-effective consistent cache.

DBOS: three years later

The VLDB Journal · 2025-04-29

Reasoning Models Can Be Effective Without Thinking

ArXiv.org · 2025-04-14

Recent LLMs have significantly improved reasoning capabilities, primarily by including an explicit, lengthy Thinking process as part of generation. In this paper, we question whether this explicit thinking is necessary. Using the state-of-the-art DeepSeek-R1-Distill-Qwen, we find that bypassing the thinking process via simple prompting, denoted as NoThinking, can be surprisingly effective. When controlling for the number of tokens, NoThinking outperforms Thinking across a diverse set of seven challenging reasoning datasets--including mathematical problem solving, formal theorem proving, and coding--especially in low-budget settings, e.g., 51.3 vs. 28.9 on ACM 23 with 700 tokens. Notably, the performance of NoThinking becomes more competitive with pass@k as k increases. Building on this observation, we demonstrate that a parallel scaling approach that uses NoThinking to generate N outputs independently and aggregates them is highly effective. For aggregation, we use task-specific verifiers when available, or we apply simple best-of-N strategies such as confidence-based selection. Our method outperforms a range of baselines with similar latency using Thinking, and is comparable to Thinking with significantly longer latency (up to 9x). Together, our research encourages a reconsideration of the necessity of lengthy thinking processes, while also establishing a competitive reference for achieving strong reasoning performance in low-budget settings or at low latency using parallel scaling.

WARP: An Efficient Engine for Multi-Vector Retrieval

2025-07-13 · 5 citations

Multi-vector retrieval methods such as ColBERT and its recent variant, the ConteXtualized Token Retriever (XTR), offer high accuracy but face efficiency challenges at scale. To address this, we present WARP, a retrieval engine that substantially improves the efficiency of retrievers trained with the XTR objective through three key innovations: (1) WARPSELECT for dynamic similarity imputation; (2) implicit decompression, avoiding costly vector reconstruction during retrieval; and (3) a two-stage reduction process for efficient score aggregation. Combined with highly-optimized C++ kernels, our system reduces end-to-end latency compared to XTR's reference implementation by 41x, and achieves a 3x speedup over the ColBERTv2/PLAID engine, while preserving retrieval quality. WARP also reduces index sizes by a factor of 2x-4x compared to XTR, enabling deployment on memory-constrained devices.

Let the Barbarians In: How AI Can Accelerate Systems Performance Research

ArXiv.org · 2025-12-16

Artificial Intelligence (AI) is beginning to transform the research process by automating the discovery of new solutions. This shift depends on the availability of reliable verifiers, which AI-driven approaches require to validate candidate solutions. Research focused on improving systems performance is especially well-suited to this paradigm because system performance problems naturally admit such verifiers: candidates can be implemented in real systems or simulators and evaluated against predefined workloads. We term this iterative cycle of generation, evaluation, and refinement AI-Driven Research for Systems (ADRS). Using several open-source ADRS instances (i.e., OpenEvolve, GEPA, and ShinkaEvolve), we demonstrate across ten case studies (e.g., multi-region cloud scheduling, mixture-of-experts load balancing, LLM-based SQL, transaction scheduling) that ADRS-generated solutions can match or even outperform human state-of-the-art designs. Based on these findings, we outline best practices (e.g., level of prompt specification, amount of feedback, robust evaluation) for effectively using ADRS, and we discuss future research directions and their implications. Although we do not yet have a universal recipe for applying ADRS across all of systems research, we hope our preliminary findings, together with the challenges we identify, offer meaningful guidance for future work as researcher effort shifts increasingly toward problem formulation and strategic oversight. Note: This paper is an extension of our prior work [14]. It adds extensive evaluation across multiple ADRS frameworks and provides deeper analysis and insights into best practices.

vAttention: Verified Sparse Attention

ArXiv.org · 2025-10-07

State-of-the-art sparse attention methods for reducing decoding latency fall into two main categories: approximate top-$k$ (and its extension, top-$p$) and recently introduced sampling-based estimation. However, these approaches are fundamentally limited in their ability to approximate full attention: they fail to provide consistent approximations across heads and query vectors and, most critically, lack guarantees on approximation quality, limiting their practical deployment. We observe that top-$k$ and random sampling are complementary: top-$k$ performs well when attention scores are dominated by a few tokens, whereas random sampling provides better estimates when attention scores are relatively uniform. Building on this insight and leveraging the statistical guarantees of sampling, we introduce vAttention, the first practical sparse attention mechanism with user-specified $(ε, δ)$ guarantees on approximation accuracy (thus, verified). These guarantees make vAttention a compelling step toward practical, reliable deployment of sparse attention at scale. By unifying top-k and sampling, vAttention outperforms both individually, delivering a superior quality-efficiency trade-off. Our experiments show that vAttention significantly improves the quality of sparse attention (e.g., $\sim$4.5 percentage points for Llama-3.1-8B-Inst and Deepseek-R1-Distill-Llama-8B on RULER-HARD), and effectively bridges the gap between full and sparse attention (e.g., across datasets, it matches full model quality with upto 20x sparsity). We also demonstrate that it can be deployed in reasoning scenarios to achieve fast decoding without compromising model quality (e.g., vAttention achieves full model quality on AIME2024 at 10x sparsity with up to 32K token generations). Code is open-sourced at https://github.com/xAlg-ai/sparse-attention-hub.

Unity Catalog: Open and Universal Governance for the Lakehouse and Beyond

2025-06-17 · 3 citations

vCache: Verified Semantic Prompt Caching

ArXiv.org · 2025-02-06

Semantic caches return cached responses for semantically similar prompts to reduce LLM inference latency and cost. They embed cached prompts and store them alongside their response in a vector database. Embedding similarity metrics assign a numerical score to quantify the similarity between a request and its nearest neighbor prompt from the cache. Existing systems use the same static similarity threshold across all requests to determine whether two prompts can share similar responses. However, we observe that static thresholds do not give formal correctness guarantees, result in unexpected error rates, and lead to suboptimal cache hit rates. This paper proposes vCache, the first verified semantic cache with user-defined error rate guarantees for predictable performance. It employs an online learning algorithm to estimate an optimal threshold for each cached prompt, enabling reliable cache responses without additional training. Our experiments show that vCache consistently meets the specified error bounds while outperforming state-of-the-art static-threshold and fine-tuned embedding baselines with up to 12.5$\times$ higher cache hit and 26$\times$ lower error rates. We release the vCache implementation and four benchmarks to support future research.

MoE-L <scp>ightning</scp> : High-Throughput MoE Inference on Memory-constrained GPUs

2025-02-06 · 4 citations

Efficient deployment of large language models, particularly Mixture of Experts (MoE) models, on resource-constrained platforms presents significant challenges in terms of computational efficiency and memory utilization. The MoE architecture, renowned for its ability to increase model capacity without a proportional increase in inference cost, greatly reduces the token generation latency compared with dense models. However, the large model size makes MoE models inaccessible to individuals without high-end GPUs. In this paper, we propose a high-throughput MoE batch inference system, MoE-Lightning, that significantly outperforms past work. MoE-Lightning introduces a novel CPU-GPU-I/O pipelining schedule, CGOPipe, with paged weights to achieve high resource utilization, and a performance model, HRM, based on a Hierarchical Roofline Model we introduce to help find policies with higher throughput than existing systems. MoE-Lightning can achieve up to (10.3x) higher throughput than state-of-the-art offloading-enabled LLM inference systems for Mixtral 8x7B on a single T4 GPU (16GB). When the theoretical system throughput is bounded by the GPU memory, MoE-Lightning can reach the throughput upper bound with 2-3x less CPU memory, significantly increasing resource utilization. MoE-Lightning also supports efficient batch inference for much larger MoEs (e.g., Mixtral 8x22B and DBRX) on multiple low-cost GPUs (e.g., 2--4 T4s).