This paper introduces rStar, a self-play mutual reasoning approach that significantly improves reasoning capabilities of small language models (SLMs) without fine-tuning or superior models. rStar decouples reasoning into a self-play mutual generation-discrimination process. First, a target SLM augments the Monte Carlo Tree Search (MCTS) with a rich set of human-like reasoning actions to construct higher quality reasoning trajectories. Next, another SLM, with capabilities similar to the target SLM, acts as a discriminator to verify each trajectory generated by the target SLM. The mutually agreed reasoning trajectories are considered mutual consistent, thus are more likely to be correct. Extensive experiments across five SLMs demonstrate rStar can effectively solve diverse reasoning problems, including GSM8K, GSM-Hard, MATH, SVAMP, and StrategyQA. Remarkably, rStar boosts GSM8K accuracy from 12.51\% to 63.91\% for LLaMA2-7B, from 36.46\% to 81.88\% for Mistral-7B, from 74.53\% to 91.13\% for LLaMA3-8B-Instruct. Code is available at https://github.com/zhentingqi/rStar.

We present rStar-Math to demonstrate that small language models (SLMs) can rival or even surpass the math reasoning capability of OpenAI o1, without distillation from superior models. rStar-Math achieves this by exercising “deep thinking” through Monte Carlo Tree Search (MCTS), where a math policy SLM performs test-time search guided by an SLM-based process reward model. rStar-Math introduces three innovations to tackle the challenges in training the two SLMs: (1) a novel code-augmented CoT data synthesis method, which performs extensive MCTS rollouts to generate step-by-step verified reasoning trajectories used to train the policy SLM; (2) a novel process reward model training method that avoids naïve step-level score annotation, yielding a more effective process preference model (PPM); (3) a self-evolution recipe in which the policy SLM and PPM are built from scratch and iteratively evolved to improve reasoning capabilities. Through 4 rounds of self-evolution with millions of synthesized solutions for 747k math problems, rStar-Math boosts SLMs’ math reasoning to state-of-the-art levels. On MATH benchmark, it improves Qwen2.5-Math-7B from 58.8% to 90.0%, surpassing o1-preview by +4.5%. On the USA Math Olympiad (AIME), rStar-Math solves an average of 53.3% (8/15) of problems, ranking among the top 20% of the brightest high school math students. Code and data are available at https://github.com/microsoft/rStar.

LongRoPE2 is a novel approach that extends the effective context window of pre-trained large language models (LLMs) to the target length, while preserving the performance on the original shorter context window. This is achieved by three contributions: (1) a hypothesis that insufficient training in higher RoPE dimensions contributes to the persistent out-of-distribution (OOD) issues observed in existing methods; (2) an effective RoPE rescaling algorithm that adopts evolutionary search guided by "needle-driven" perplexity to address the insufficient training problem; (3) a mixed context window training approach that fine-tunes model weights to adopt rescaled RoPE for long-context sequences while preserving the short-context performance with the original RoPE. Extensive experiments on LLaMA3-8B and Phi3-mini-3.8B across various benchmarks validate the hypothesis and demonstrate the effectiveness of LongRoPE2. Remarkably, LongRoPE2 extends LLaMA3-8B to achieve a 128K effective context length while retaining over 98.5% of short-context performance, using only 10B tokens – 80x fewer than Meta’s approach, which fails to reach the target effective context length.

Advancing code reasoning in large language models (LLMs) is fundamentally limited by the scarcity of high-difficulty datasets, especially those with verifiable input-output test cases necessary for rigorous solution validation at scale. We introduce rStar-Coder, which significantly improves LLM code reasoning capabilities by constructing a large-scale, verified dataset of 418K competition-level code problems, 580K long-reasoning solutions along with rich test cases of varying difficulty. This is achieved through three core contributions: (1) we curate competitive programming code problems and solutions to synthesize new, solvable problems; (2) we introduce a reliable input-output test case synthesis pipeline that decouples the generation into a three-step input generation method and a mutual verification mechanism for effective output labeling; (3) we augment problems with high-quality, test-case-verified long-reasoning solutions. Extensive experiments on Qwen models (1.5B-14B) across various code reasoning benchmarks demonstrate the superiority of rStar-Coder dataset, achieving leading performance comparable to frontier reasoning LLMs with significantly smaller model sizes. On LiveCodeBench, rStar-Coder improves Qwen2.5-7B from 17.4% to an impressive 57.3%, and Qwen2.5-14B from 23.3% to 62.5%, surpassing o3-mini (low) by 3.1%. On the more challenging USA Computing Olympiad, our 7B model achieves an average pass@1 accuracy of 16.15%, outperforming the frontier-level QWQ-32B. rStar-Coder dataset is publicly available at https://huggingface.co/datasets/microsoft/rStar-Coder.

Large Language Models (LLMs) have shown impressive capabilities, yet they still struggle with math reasoning.In this work, we propose CoT-Influx, a novel approach that pushes the boundary of few-shot Chain-of-Thoughts (CoT) learning to improve LLM mathematical reasoning.Motivated by the observation that adding more concise CoT examples in the prompt can improve LLM reasoning performance, CoT-Influx employs a coarse-to-fine pruner to maximize the input of effective and concise CoT examples.The pruner first selects as many crucial CoT examples as possible and then prunes unimportant tokens to fit the context window.A math reasoning dataset with diverse difficulty levels and reasoning steps is used to train the pruner, along with a math-specialized reinforcement learning approach.As a result, by enabling more CoT examples with double the context window size in tokens, CoT-Influx significantly outperforms various prompting baselines across various LLMs (LLaMA2-7B, 13B, 70B) and 6 math datasets, achieving up to 4.40% absolute improvements.Remarkably, without any fine-tuning, LLaMA2-70B with CoT-Influx surpasses GPT-3.5 and a wide range of larger LLMs (PaLM, Minerva 540B, etc.) on GSM8K.CoT-Influx is a plug-and-play module for LLMs, adaptable in various scenarios.It's compatible with advanced reasoning prompting techniques, such as self-consistency, and supports different long-context LLMs, including Mistral-7B-v0.3-32K and Yi-6B-200K.Codes are available at https://github.com/HuangOwen/CoT-Influx

Scaling model size significantly challenges the deployment and inference of Large Language Models (LLMs).Due to the redundancy in LLM weights, recent research has focused on pushing weight-only quantization to extremely low-bit (even down to 2 bits).It reduces memory requirements, optimizes storage costs, and * Contribution during internship at Microsoft Research

Large context window is a desirable feature in large language models (LLMs). However, due to high fine-tuning costs, scarcity of long texts, and catastrophic values introduced by new token positions, current extended context windows are limited to around 128k tokens. This paper introduces LongRoPE that, for the first time, extends the context window of pre-trained LLMs to an impressive 2048k tokens, with up to only 1k fine-tuning steps at within 256k training lengths, while maintaining performance at the original short context window. This is achieved by three key innovations: (i) we identify and exploit two forms of non-uniformities in positional interpolation through an efficient search, providing a better initialization for fine-tuning and enabling an 8x extension in non-fine-tuning scenarios; (ii) we introduce a progressive extension strategy that first fine-tunes a 256k length LLM and then conducts a second positional interpolation on the fine-tuned extended LLM to achieve a 2048k context window; (iii) we readjust LongRoPE on 8k length to recover the short context window performance. Extensive experiments on LLaMA2 and Mistral across various tasks demonstrate the effectiveness of our method. Models extended via LongRoPE retain the original architecture with minor modifications to the positional embedding, and can reuse most pre-existing optimizations. Code is available at https://github.com/microsoft/LongRoPE

Neural Architecture Search (NAS) has shown promising performance in the automatic design of vision transformers (ViT) exceeding 1G FLOPs. However, designing lightweight and low-latency ViT models for diverse mobile devices remains a big challenge. In this work, we propose ElasticViT, a two-stage NAS approach that trains a high-quality ViT supernet over a very large search space for covering a wide range of mobile devices, and then searches an optimal sub-network (subnet) for direct deployment. However, current supernet training methods that rely on uniform sampling suffer from the gradient conflict issue: the sampled subnets can have vastly different model sizes (e.g., 50M vs. 2G FLOPs), leading to different optimization directions and inferior performance. To address this challenge, we propose two novel sampling techniques: complexity-aware sampling and performance-aware sampling. Complexity-aware sampling limits the FLOPs difference among the subnets sampled across adjacent training steps, while covering different-sized subnets in the search space. Performance-aware sampling further selects subnets that have good accuracy, which can reduce gradient conflicts and improve supernet quality. Our discovered models, ElasticViT models, achieve top-1 accuracy from 67.2% to 80.0% on ImageNet from 60M to 800M FLOPs without extra retraining, outperforming all prior CNNs and ViTs in terms of accuracy and latency. Our tiny and small models are also the first ViT models that surpass state-of-the-art CNNs with significantly lower latency on mobile devices. For instance, ElasticViT-S1 runs 2.62× faster than EfficientNet-B0 with 0.1% higher accuracy.

The combination of Neural Architecture Search (NAS) and quantization has proven successful in automatically designing low-FLOPs INT8 quantized neural networks (QNN). However, directly applying NAS to design accurate QNN models that achieve low latency on real-world devices leads to inferior performance. In this work, we identify that the poor INT8 latency is due to the quantization-unfriendly issue: the operator and configuration (e.g., channel width) choices in prior art search spaces lead to diverse quantization efficiency and can slow down the INT8 inference speed. To address this challenge, we propose SpaceEvo, an automatic method for designing a dedicated, quantization-friendly search space for each target hardware. The key idea of SpaceEvo is to automatically search hardware-preferred operators and configurations to construct the search space, guided by a metric called Q-T score to quantify how quantization-friendly a candidate search space is. We further train a quantized-for-all supernet over our discovered search space, enabling the searched models to be directly deployed without extra retraining or quantization. Our discovered models, SEQnet, establish new SOTA INT8 quantized accuracy under various latency constraints, achieving up to 10.1% accuracy improvement on ImageNet than prior art CNNs under the same latency. Extensive experiments on real devices show that SpaceEvo consistently outperforms manually-designed search spaces with up to 2.5× faster speed while achieving the same accuracy.

Deploying pre-trained transformer models like BERT on downstream tasks in resource-constrained scenarios is challenging due to their high inference cost, which grows rapidly with input sequence length. In this work, we propose a constraint-aware and ranking-distilled token pruning method ToP, which selectively removes unnecessary tokens as input sequence passes through layers, allowing the model to improve online inference speed while preserving accuracy. ToP overcomes the limitation of inaccurate token importance ranking in the conventional self-attention mechanism through a ranking-distilled token distillation technique, which distills effective token rankings from the final layer of unpruned models to early layers of pruned models. Then, ToP introduces a coarse-to-fine pruning approach that automatically selects the optimal subset of transformer layers and optimizes token pruning decisions within these layers through improved L0 regularization. Extensive experiments on GLUE benchmark and SQuAD tasks demonstrate that ToP outperforms state-of-the-art token pruning and model compression methods with improved accuracy and speedups. ToP reduces the average FLOPs of BERT by 8.1X while achieving competitive accuracy on GLUE, and provides a real latency speedup of up to 7.4X on an Intel CPU. Code is available at https://github.com/microsoft/Moonlit/tree/main/ToP

On-device Deep Neural Network (DNN) inference consumes significant computing resources and development efforts. To alleviate that, we propose LUT-NN, the first system to empower inference by table lookup, to reduce inference cost. LUT-NN learns the typical features for each operator, named centroid, and precompute the results for these centroids to save in lookup tables. During inference, the results of the closest centroids with the inputs can be read directly from the table, as the approximated outputs without computations.

Ad relevance modeling plays a critical role in online advertising systems including Microsoft Bing. To leverage powerful transformers like BERT in this low-latency setting, many existing approaches perform ad-side computations offline. While efficient, these approaches are unable to serve cold start ads, resulting in poor relevance predictions for such ads. This work aims to design a new, low-latency BERT via structured pruning to empower real-time online inference for cold start ads relevance on a CPU platform. Our challenge is that previous methods typically prune all layers of the transformer to a high, uniform sparsity, thereby producing models which cannot achieve satisfactory inference speed with an acceptable accuracy.

Human brains are known to be capable of speeding up visual recognition of repeatedly presented objects through faster memory encoding and accessing procedures on activated neurons. For the first time, we borrow and distill such a capability into a semantic memory design, namely SMTM, to improve on-device CNN inference. SMTM employs a hierarchical memory architecture to leverage the long-tail distribution of objects of interest, and further incorporates several novel techniques to put it into effects: (1) it encodes high-dimensional feature maps into low-dimensional, semantic vectors for low-cost yet accurate cache and lookup; (2) it uses a novel metric in determining the exit timing considering different layers' inherent characteristics; (3) it adaptively adjusts the cache size and semantic vectors to fit the scene dynamics. SMTM is prototyped on commodity CNN engine and runs on both mobile CPU and GPU. Extensive experiments on large-scale datasets and models show that SMTM can significantly speed up the model inference over standard approach (up to 2×) and prior cache designs (up to 1.5x), with acceptable accuracy loss.

With the recent trend of on-device deep learning, inference latency has become a crucial metric in running Deep Neural Network (DNN) models on various mobile and edge devices. To this end, latency prediction of DNN model inference is highly desirable for many tasks where measuring the latency on real devices is infeasible or too costly, such as searching for efficient DNN models with latency constraints from a huge model-design space. Yet it is very challenging and existing approaches fail to achieve a high accuracy of prediction, due to the varying model-inference latency caused by the runtime optimizations on diverse edge devices.

Given the emerging concerns over app privacy-related risks, major app distribution providers (e.g., Microsoft) have been exploring approaches to help end users to make informed decision before installation. This is different from existing approaches of simply trusting users to make the right decision. We build on the direction of risk rating as the way to communicate app-specific privacy risks to end users. To this end, we propose to use sensitivity analysis to infer whether an app requests sensitive on-device resources/ data that are not required for its expected functionality. Our system, Privet, addresses challenges in efficiently achieving test coverage and automated privacy risk assessment. Finally, we evaluate Privet with 1,000 Android apps released in the wild.

Contrary to popular belief, mobile apps can spend a large fraction of time running "hidden" as background services. And, bugs in services can translate into crashes, energy depletion, device slow-down, etc. Unfortunately, without necessary testing tools, developers can only resort to telemetries from user devices in the wild. To this end, Snowdrop is a testing framework that systematically identifies and automates background services in Android apps. Snowdrop realizes a service-oriented approach that does not assume all inter-component communication messages are explicitly coded in the app bytecode. Furthermore, to improve the completeness of test inputs generated, Snowdrop infers field values by exploiting the similarity in how developers name variables. We evaluate Snowdrop by testing 848 commercially available mobile apps. Empirical results show that Snowdrop can achieve 20.91% more code path coverage than pathwise test input generators, and 64.11% more coverage than random test input generators.