Siddhartha Jain

Ben Athiwaratkun, Sanjay Krishna Gouda, Zijian Wang et al. · amazon-science, ibm-research

We present new benchmarks on evaluation code generation models: MBXP and Multilingual HumanEval, and MathQA-X. These datasets cover over 10 programming languages and are generated using a scalable conversion framework that transpiles prompts and test cases from the original Python datasets into the corresponding data in the target language. Using these benchmarks, we are able to assess the performance of code generation models in a multi-lingual fashion, and discovered generalization ability of language models on out-of-domain languages, advantages of multi-lingual models over mono-lingual, the ability of few-shot prompting to teach the model new languages, and zero-shot translation abilities even on mono-lingual settings. Furthermore, we use our code generation model to perform large-scale bootstrapping to obtain synthetic canonical solutions in several languages, which can be used for other code-related evaluations such as code insertion, robustness, or summarization tasks. Overall, our benchmarks represents a significant step towards a deeper understanding of language models' code generation abilities. We publicly release our code and datasets at https://github.com/amazon-research/mxeval.

Aaron Blakeman, Aarti Basant, Abhinav Khattar et al. · nvidia

As inference-time scaling becomes critical for enhanced reasoning capabilities, it is increasingly becoming important to build models that are efficient to infer. We introduce Nemotron-H, a family of 8B and 56B/47B hybrid Mamba-Transformer models designed to reduce inference cost for a given accuracy level. To achieve this goal, we replace the majority of self-attention layers in the common Transformer model architecture with Mamba layers that perform constant computation and require constant memory per generated token. We show that Nemotron-H models offer either better or on-par accuracy compared to other similarly-sized state-of-the-art open-sourced Transformer models (e.g., Qwen-2.5-7B/72B and Llama-3.1-8B/70B), while being up to 3$\times$ faster at inference. To further increase inference speed and reduce the memory required at inference time, we created Nemotron-H-47B-Base from the 56B model using a new compression via pruning and distillation technique called MiniPuzzle. Nemotron-H-47B-Base achieves similar accuracy to the 56B model, but is 20% faster to infer. In addition, we introduce an FP8-based training recipe and show that it can achieve on par results with BF16-based training. This recipe is used to train the 56B model. We are releasing Nemotron-H base model checkpoints with support in Hugging Face and NeMo.

Aakshita Chandiramani, Aaron Blakeman, Abdullahi Olaoye et al. · amazon-science, cmu

We describe the pre-training, post-training, and quantization of Nemotron 3 Super, a 120 billion (active 12 billion) parameter hybrid Mamba-Attention Mixture-of-Experts model. Nemotron 3 Super is the first model in the Nemotron 3 family to 1) be pre-trained in NVFP4, 2) leverage LatentMoE, a new Mixture-of-Experts architecture that optimizes for both accuracy per FLOP and accuracy per parameter, and 3) include MTP layers for inference acceleration through native speculative decoding. We pre-trained Nemotron 3 Super on 25 trillion tokens followed by post-training using supervised fine tuning (SFT) and reinforcement learning (RL). The final model supports up to 1M context length and achieves comparable accuracy on common benchmarks, while also achieving up to 2.2x and 7.5x higher inference throughput compared to GPT-OSS-120B and Qwen3.5-122B, respectively. Nemotron 3 Super datasets, along with the base, post-trained, and quantized checkpoints, are open-sourced on HuggingFace.

Aarti Basant, Abhijit Khairnar, Abhijit Paithankar et al. · nvidia

We introduce Nemotron-Nano-9B-v2, a hybrid Mamba-Transformer language model designed to increase throughput for reasoning workloads while achieving state-of-the-art accuracy compared to similarly-sized models. Nemotron-Nano-9B-v2 builds on the Nemotron-H architecture, in which the majority of the self-attention layers in the common Transformer architecture are replaced with Mamba-2 layers, to achieve improved inference speed when generating the long thinking traces needed for reasoning. We create Nemotron-Nano-9B-v2 by first pre-training a 12-billion-parameter model (Nemotron-Nano-12B-v2-Base) on 20 trillion tokens using an FP8 training recipe. After aligning Nemotron-Nano-12B-v2-Base, we employ the Minitron strategy to compress and distill the model with the goal of enabling inference on up to 128k tokens on a single NVIDIA A10G GPU (22GiB of memory, bfloat16 precision). Compared to existing similarly-sized models (e.g., Qwen3-8B), we show that Nemotron-Nano-9B-v2 achieves on-par or better accuracy on reasoning benchmarks while achieving up to 6x higher inference throughput in reasoning settings like 8k input and 16k output tokens. We are releasing Nemotron-Nano-9B-v2, Nemotron-Nano12B-v2-Base, and Nemotron-Nano-9B-v2-Base checkpoints along with the majority of our pre- and post-training datasets on Hugging Face.

Siddhartha Jain, Xiaofei Ma, Anoop Deoras et al. · amazon-science

Large Language Models (LLMs) can exhibit considerable variation in the quality of their sampled outputs. Reranking and selecting the best generation from the sampled set is a popular way of obtaining strong gains in generation quality. In this paper, we present a novel approach for reranking LLM generations. Unlike other techniques that might involve additional inferences or training a specialized reranker, our approach relies on easy to compute pairwise statistics between the generations that have minimal compute overhead. We show that our approach can be formalized as an extension of self-consistency and analyze its performance in that framework, theoretically as well as via simulations. We show strong improvements for selecting the best k generations for code generation tasks as well as robust improvements for the best generation for the tasks of autoformalization, summarization, and translation. While our approach only assumes black-box access to LLMs, we show that additional access to token probabilities can improve performance even further.

Gabriel Ryan, Siddhartha Jain, Mingyue Shang et al.

Testing plays a pivotal role in ensuring software quality, yet conventional Search Based Software Testing (SBST) methods often struggle with complex software units, achieving suboptimal test coverage. Recent works using large language models (LLMs) for test generation have focused on improving generation quality through optimizing the test generation context and correcting errors in model outputs, but use fixed prompting strategies that prompt the model to generate tests without additional guidance. As a result LLM-generated testsuites still suffer from low coverage. In this paper, we present SymPrompt, a code-aware prompting strategy for LLMs in test generation. SymPrompt's approach is based on recent work that demonstrates LLMs can solve more complex logical problems when prompted to reason about the problem in a multi-step fashion. We apply this methodology to test generation by deconstructing the testsuite generation process into a multi-stage sequence, each of which is driven by a specific prompt aligned with the execution paths of the method under test, and exposing relevant type and dependency focal context to the model. Our approach enables pretrained LLMs to generate more complete test cases without any additional training. We implement SymPrompt using the TreeSitter parsing framework and evaluate on a benchmark challenging methods from open source Python projects. SymPrompt enhances correct test generations by a factor of 5 and bolsters relative coverage by 26% for CodeGen2. Notably, when applied to GPT-4, SymPrompt improves coverage by over 2x compared to baseline prompting strategies.

Akhiad Bercovich, Itay Levy, Izik Golan et al. · nvidia

We introduce the Llama-Nemotron series of models, an open family of heterogeneous reasoning models that deliver exceptional reasoning capabilities, inference efficiency, and an open license for enterprise use. The family comes in three sizes -- Nano (8B), Super (49B), and Ultra (253B) -- and performs competitively with state-of-the-art reasoning models such as DeepSeek-R1 while offering superior inference throughput and memory efficiency. In this report, we discuss the training procedure for these models, which entails using neural architecture search from Llama 3 models for accelerated inference, knowledge distillation, and continued pretraining, followed by a reasoning-focused post-training stage consisting of two main parts: supervised fine-tuning and large scale reinforcement learning. Llama-Nemotron models are the first open-source models to support a dynamic reasoning toggle, allowing users to switch between standard chat and reasoning modes during inference. To further support open research and facilitate model development, we provide the following resources: 1. We release the Llama-Nemotron reasoning models -- LN-Nano, LN-Super, and LN-Ultra -- under the commercially permissive NVIDIA Open Model License Agreement. 2. We release the complete post-training dataset: Llama-Nemotron-Post-Training-Dataset. 3. We also release our training codebases: NeMo, NeMo-Aligner, and Megatron-LM.

Wasi Uddin Ahmad, Sean Narenthiran, Somshubra Majumdar et al.

Since the advent of reasoning-based large language models, many have found great success from distilling reasoning capabilities into student models. Such techniques have significantly bridged the gap between reasoning and standard LLMs on coding tasks. Despite this, much of the progress on distilling reasoning models remains locked behind proprietary datasets or lacks details on data curation, filtering and subsequent training. To address this, we construct a superior supervised fine-tuning (SFT) dataset that we use to achieve state-of-the-art coding capability results in models of various sizes. Our distilled models use only SFT to achieve 61.8% on LiveCodeBench and 24.6% on CodeContests, surpassing alternatives trained with reinforcement learning. We then perform analysis on the data sources used to construct our dataset, the impact of code execution filtering, and the importance of instruction/solution diversity. We observe that execution filtering negatively affected benchmark accuracy, leading us to prioritize instruction diversity over solution correctness. Finally, we also analyze the token efficiency and reasoning patterns utilized by these models. We will open-source these datasets and distilled models to the community.

Shubham Toshniwal, Aleksander Ficek, Siddhartha Jain et al.

Scaling test-time compute via parallel sampling can substantially improve LLM reasoning, but is often limited by Best-of-N selection quality. Generative selection methods, such as GenSelect, address this bottleneck, yet strong selection performance remains largely limited to large models. We show that small reasoning models can acquire strong GenSelect capabilities through targeted reinforcement learning. To this end, we synthesize selection tasks from large-scale math and code instruction datasets by filtering to instances with both correct and incorrect candidate solutions, and train 1.7B-parameter models with DAPO to reward correct selections. Across math (AIME24, AIME25, HMMT25) and code (LiveCodeBench) reasoning benchmarks, our models consistently outperform prompting and majority-voting baselines, often approaching or exceeding much larger models. Moreover, these gains generalize to selecting outputs from stronger models despite training only on outputs from weaker models. Overall, our results establish reinforcement learning as a scalable way to unlock strong generative selection in small models, enabling efficient test-time scaling.

Wasi Uddin Ahmad, Somshubra Majumdar, Aleksander Ficek et al.

Recent advancements in reasoning-based Large Language Models (LLMs), particularly their potential through test-time scaling, have created significant opportunities for distillation in code generation and critique. However, progress in both areas fundamentally depends on large-scale, high-quality datasets. In this work, we introduce OpenCodeReasoning-II, a dataset consists of 2.5M question-solution-critique triples (approx. 35K unique programming questions), making it nearly twice the size of the previous largest publicly available code reasoning dataset. In this work, we employ a two-stage supervised fine-tuning strategy. The first stage focuses on fine-tuning for code generation, while the second stage involves the joint training of models for both code generation and critique. Our resulting finetuned Qwen2.5-Instruct models achieve performance in code generation that either exceeds or equals the best prior open-weight distilled models. Notably, the integration of our code generation and critique models leads to significant improvements in competitive coding performance. Furthermore, we present an extension of the LiveCodeBench benchmark to specifically support the C++ programming language, thereby facilitating more comprehensive LLM evaluation using this benchmark.

Colin White, Samuel Dooley, Manley Roberts et al.

Test set contamination, wherein test data from a benchmark ends up in a newer model's training set, is a well-documented obstacle for fair LLM evaluation and can quickly render benchmarks obsolete. To mitigate this, many recent benchmarks crowdsource new prompts and evaluations from human or LLM judges; however, these can introduce significant biases, and break down when scoring hard questions. In this work, we introduce a new benchmark for LLMs designed to be resistant to both test set contamination and the pitfalls of LLM judging and human crowdsourcing. We release LiveBench, the first benchmark that (1) contains frequently-updated questions from recent information sources, (2) scores answers automatically according to objective ground-truth values, and (3) contains a wide variety of challenging tasks, spanning math, coding, reasoning, language, instruction following, and data analysis. To achieve this, LiveBench contains questions that are based on recently-released math competitions, arXiv papers, news articles, and datasets, and it contains harder, contamination-limited versions of tasks from previous benchmarks such as Big-Bench Hard, AMPS, and IFEval. We evaluate many prominent closed-source models, as well as dozens of open-source models ranging from 0.5B to 405B in size. LiveBench is difficult, with top models achieving below 70% accuracy. We release all questions, code, and model answers. Questions are added and updated on a monthly basis, and we release new tasks and harder versions of tasks over time so that LiveBench can distinguish between the capabilities of LLMs as they improve in the future. We welcome community engagement and collaboration for expanding the benchmark tasks and models.

Mehrzad Samadi, Aleksander Ficek, Sean Narenthiran et al.

Competitive programming has become a rigorous benchmark for evaluating the reasoning and problem-solving capabilities of large language models (LLMs). The International Olympiad in Informatics (IOI) stands out as one of the most prestigious annual competitions in competitive programming and has become a key benchmark for comparing human and AI-level programming ability. While several proprietary models have been claimed to achieve gold medal-level performance at the IOI, often with undisclosed methods, achieving comparable results with open-weight models remains a significant challenge. In this paper, we present \gencluster, a scalable and reproducible test-time compute framework that attains IOI gold-level performance using open-weight models. It combines large-scale generation, behavioral clustering, ranking, and a round-robin submission strategy to efficiently explore diverse solution spaces under limited validation budgets. Our experiments show that the performance of our proposed approach scales consistently with available compute, narrowing the gap between open and closed systems. Notably, we will show that GenCluster can achieve a gold medal at IOI 2025 for the first time with an open-weight model gpt-oss-120b, setting a new benchmark for transparent and reproducible evaluation of reasoning in LLMs.

Junlin Wang, Siddhartha Jain, Dejiao Zhang et al.

A diverse array of reasoning strategies has been proposed to elicit the capabilities of large language models. However, in this paper, we point out that traditional evaluations which focus solely on performance metrics miss a key factor: the increased effectiveness due to additional compute. By overlooking this aspect, a skewed view of strategy efficiency is often presented. This paper introduces a framework that incorporates the compute budget into the evaluation, providing a more informative comparison that takes into account both performance metrics and computational cost. In this budget-aware perspective, we find that complex reasoning strategies often don't surpass simpler baselines purely due to algorithmic ingenuity, but rather due to the larger computational resources allocated. When we provide a simple baseline like chain-of-thought self-consistency with comparable compute resources, it frequently outperforms reasoning strategies proposed in the literature. In this scale-aware perspective, we find that unlike self-consistency, certain strategies such as multi-agent debate or Reflexion can become worse if more compute budget is utilized.

Brandon Carter, Siddhartha Jain, Jonas Mueller et al.

Image classifiers are typically scored on their test set accuracy, but high accuracy can mask a subtle type of model failure. We find that high scoring convolutional neural networks (CNNs) on popular benchmarks exhibit troubling pathologies that allow them to display high accuracy even in the absence of semantically salient features. When a model provides a high-confidence decision without salient supporting input features, we say the classifier has overinterpreted its input, finding too much class-evidence in patterns that appear nonsensical to humans. Here, we demonstrate that neural networks trained on CIFAR-10 and ImageNet suffer from overinterpretation, and we find models on CIFAR-10 make confident predictions even when 95% of input images are masked and humans cannot discern salient features in the remaining pixel-subsets. We introduce Batched Gradient SIS, a new method for discovering sufficient input subsets for complex datasets, and use this method to show the sufficiency of border pixels in ImageNet for training and testing. Although these patterns portend potential model fragility in real-world deployment, they are in fact valid statistical patterns of the benchmark that alone suffice to attain high test accuracy. Unlike adversarial examples, overinterpretation relies upon unmodified image pixels. We find ensembling and input dropout can each help mitigate overinterpretation.

Siddhartha Jain, Ge Liu, David Gifford

We introduce Information Condensing Active Learning (ICAL), a batch mode model agnostic Active Learning (AL) method targeted at Deep Bayesian Active Learning that focuses on acquiring labels for points which have as much information as possible about the still unacquired points. ICAL uses the Hilbert Schmidt Independence Criterion (HSIC) to measure the strength of the dependency between a candidate batch of points and the unlabeled set. We develop key optimizations that allow us to scale our method to large unlabeled sets. We show significant improvements in terms of model accuracy and negative log likelihood (NLL) on several image datasets compared to state of the art batch mode AL methods for deep learning.

Siddhartha Jain, Ge Liu, Jonas Mueller et al.

The inaccuracy of neural network models on inputs that do not stem from the training data distribution is both problematic and at times unrecognized. Model uncertainty estimation can address this issue, where uncertainty estimates are often based on the variation in predictions produced by a diverse ensemble of models applied to the same input. Here we describe Maximize Overall Diversity (MOD), a straightforward approach to improve ensemble-based uncertainty estimates by encouraging larger overall diversity in ensemble predictions across all possible inputs that might be encountered in the future. When applied to various neural network ensembles, MOD significantly improves predictive performance for out-of-distribution test examples without sacrificing in-distribution performance on 38 Protein-DNA binding regression datasets, 9 UCI datasets, and the IMDB-Wiki image dataset. Across many Bayesian optimization tasks, the performance of UCB acquisition is also greatly improved by leveraging MOD uncertainty estimates.

Brandon Carter, Jonas Mueller, Siddhartha Jain et al.

Local explanation frameworks aim to rationalize particular decisions made by a black-box prediction model. Existing techniques are often restricted to a specific type of predictor or based on input saliency, which may be undesirably sensitive to factors unrelated to the model's decision making process. We instead propose sufficient input subsets that identify minimal subsets of features whose observed values alone suffice for the same decision to be reached, even if all other input feature values are missing. General principles that globally govern a model's decision-making can also be revealed by searching for clusters of such input patterns across many data points. Our approach is conceptually straightforward, entirely model-agnostic, simply implemented using instance-wise backward selection, and able to produce more concise rationales than existing techniques. We demonstrate the utility of our interpretation method on various neural network models trained on text, image, and genomic data.