While large language models (LLMs) demonstrate impressive capabilities across numerous applications, their robustness remains a critical concern. This paper is motivated by a specific vulnerability: the order sensitivity of LLMs. This vulnerability manifests itself as the order bias observed when LLMs decide between possible options (for example, a preference for the first option) and the tendency of LLMs to provide different answers when options are reordered. The use cases for this scenario extend beyond the classical case of multiple-choice question answering to the use of LLMs for multidocument tasks and as automated evaluators in AI pipelines. We introduce Set-LLM, a novel architectural adaptation for pretrained LLMs that enables the processing of mixed set-text inputs with permutation invariance guarantees. The adaptations involve a new attention mask and new positional encodings specifically designed for sets. We provide a theoretical proof of invariance and demonstrate through experiments that Set-LLM can be trained effectively, achieving comparable or improved performance and maintaining the runtime of the original model, while altogether eliminating order sensitivity.

For Large Language Models (LLMs) to be reliably deployed in both everyday and high-stakes domains, knowing when not to answer is equally critical as answering correctly. Real-world user queries, which can be underspecified, ill-posed, or fundamentally unanswerable, require LLMs to reason about uncertainty and selectively abstain--i.e., refuse to answer definitively. However, abstention remains understudied, without a systematic evaluation framework for modern LLMs. In this work, we introduce AbstentionBench: a large-scale benchmark for holistically evaluating abstention across 20 diverse datasets, including questions with unknown answers, underspecification, false premises, subjective interpretations, and outdated information. Evaluating 20 frontier LLMs reveals abstention is an unsolved problem, and one where scaling models is of little use. While recent reasoning LLMs have shown impressive results in complex problem solving, surprisingly, we find that reasoning fine-tuning degrades abstention (by 24% on average), even for math and science domains on which reasoning models are explicitly trained. We find that while a carefully crafted system prompt can boost abstention in practice, it does not resolve models' fundamental inability to reason about uncertainty. We release AbstentionBenchto foster research into advancing LLM reliability.2

Generating diverse responses from large language models (LLMs) is crucial for applications such as adversarial testing, search, and synthetic data generation, where diversity provides distinct answers across generations. Previous approaches rely solely on increasing the temperature, sacrificing quality. Furthermore, the model's next-token probabilities may not be representative of the true answer distribution. To combat these challenges, we propose SimpleStrat, an alternative that uses the language sample. To model measure itself resampling to partition divers the ity solution, we introduce space int Co o verageQA, strata from a dataset which of to underspecified questions with multiple equally plausible answers. We propose measuring resampling diversity as the KLDivergence between the response distribution and the uniform distribution over valid ground truth answers and use recall as an alternative when assessing proprietary models. On CoverageQA, SimpleStrat improves diversity across all temperatures, showing orthogonal benefits. Quantifiably, we achieve as much as 4X better recall when applied to GPT-4o, and an average reLineduction in KL divergence by 0.36 when applied to Llama 3. Furthermore, we showthat SimpleStrat achieves more resampling diversity at temperature T=0 than scaling and temperature dataset available to T=1 at on https://github.com/j

The primary goal of PACBench is to catalyze the development of more capable, reliable, and physically grounded VLMs and their fine-tuned variants, often called VLAs for real-world robotic applications. Because VLA fine-tuning typically relies on low-level trajectory data rather than higher level reasoning, probing the underlying VLM's understanding of object Properties, action Affordances, and physical Constraints (PAC) gives us a grounded lens into the capabilities that downstream robotic policies will inherit. By diagnosing PAC weaknesses in the base model, researchers can distinguish whether a VLA's performance stems from genuine physical common sense or simply memorized motion patterns, and thus guide targeted improvements in model architectures, training methodologies, and dataset curation. In doing so, PACBench helps ensure that robotic systems become more predictable, less prone to errors from a lack of physical understanding, and better equipped for safe, effective collaboration in complex, everyday environments. By providing a fine-grained diagnostic tool, PACBench can help researchers and developers identify specific weaknesses in current models, thereby guiding targeted improvements in model architectures, training methodologies, and dataset curation. This, in turn, can lead to robotic systems that are more predictable, less prone to errors stemming from a lack of physical common sense, and better able to perform a wide range of useful tasks. The open release of our benchmark and its diverse data sources (including web-scale images, real-world humanoid captures, and simulated scenarios) is intended to foster broad community engagement and accelerate progress in this crucial area of AI. While any advancement in AI capabilities warrants ongoing consideration of its societal implications, our work focuses on enhancing the fundamental understanding and robustness of AI systems, which we see as a positive step towards more responsible AI development.

Vision-Language Models (VLMs) are increasingly pivotal for generalist robot manipulation, enabling tasks such as physical reasoning, policy generation, and failure detection. However, their proficiency in these high-level applications often assumes a deep understanding of low-level physical prerequisites, a capability that is largely unverified. To perform actions reliably, robots must comprehend intrinsic object properties (e.g., material, weight), action affordances (e.g., graspable, stackable), and physical constraints (e.g., stability, reachability, or an object's state like being closed). Despite their ubiquitous use in manipulation, we argue that off-the-shelf VLMs may lack this granular, physically-grounded understanding, as these specific prerequisites are often overlooked during training. Addressing this critical gap, we introduce PACBench, a comprehensive benchmark designed to systematically evaluate VLMs on their understanding of these core Properties, Affordances, and Constraints (PAC) from a task executability perspective. PAC Bench features a diverse dataset with more than 30,000 annotations, comprising 673 real-world images (115 object classes, 15 property types, 1-3 affordances defined per object class), 100 real-world humanoid view scenarios, and 120 unique simulated constraint scenarios across four tasks. Our evaluations reveal significant gaps in the ability of VLMs to grasp fundamental physical concepts, underscoring their current limitations for reliable robot manipulation and pointing to key areas that require targeted research. PACBench also serves as a standardized benchmark for rigorously evaluating the physical reasoning capabilities of VLMs guiding the development of more robust and physically grounded models for robot manipulation.

As the adoption of Generative AI in real-world services grow explosively, energy has emerged as a critical bottleneck resource. However, energy remains a metric that is often overlooked, under-explored, or poorly understood in the context of building ML systems. We present the ML.ENERGY Benchmark, a benchmark suite and tool for measuring inference energy consumption under realistic service environments, and the corresponding ML.ENERGY Leaderboard, which have served as a valuable resource for those hoping to understand and optimize the energy consumption of their generative AI services. In this paper, we explain four key design principles for benchmarking ML energy we have acquired over time, and then describe how they are implemented in the ML.ENERGY Benchmark. We then highlight results from the early 2025 iteration of the benchmark, including energy measurements of 40 widely used model architectures across 6 different tasks, case studies of how ML design choices impact energy consumption, and how automated optimization recommendations can lead to significant (sometimes more than 40%) energy savings without changing what is being computed by the model. The ML.ENERGY Benchmark is open-source and can be easily extended to various customized models and application scenarios.

Large Language Models (LLMs) have demonstrated impressive capabilities as intelligent agents capable of solving complex problems.

Large-scale AI models, such as Large Language Models (LLMs) and Diffusion Models (DMs), have grown rapidly in size, creating significant challenges for efficient deployment on resource-constrained hardware. In this paper, we introduce Dynamic-Length Float (DFloat11), a lossless compression framework that reduces LLM and DM size by 30% while preserving outputs that are bit-for-bit identical to the original model. DFloat11 is motivated by the low entropy in the BFloat16 weight representation of LLMs, which reveals significant inefficiency in the existing storage format. By applying entropy coding, DFloat11 assigns dynamic-length encodings to weights based on frequency, achieving near information-optimal compression without any loss of precision. To facilitate efficient inference with dynamic-length encodings, we develop a custom GPU kernel for fast online decompression. Our design incorporates the following: (i) compact, hierarchical lookup tables (LUTs) that fit within GPUSRAM for efficient decoding, (ii) a two-phase GPU kernel for coordinating thread read/write positions using lightweight auxiliary variables, and (iii) transformer-block-level decompression to minimize latency. Experiments on Llama 3.3, Qwen 3, Mistral 3, FLUX.1, and others validate our hypothesis that DFloat11 achieves around 30% model size reduction while preserving bit-for-bit identical outputs. Compared to a potential alternative of offloading parts of an uncompressed model to the CPU to meet memory constraints, DFloat11 achieves 2.3-46.2

We introduce AutoJudge1, a method that accelerates large language model (LLM) inference with task-specific lossy speculative decoding. Instead of matching the original model output distribution token-by-token, we identify the generated tokens that affect the downstream quality of the response, relaxing the distribution match guarantee so that the "unimportant" tokens can be generated faster. Our approach relies on a semi-greedy search algorithm to test which of the mismatches between target and draft models should be corrected to preserve quality and which ones may be skipped. We then train a lightweight classifier based on existing LLM embeddings to predict, at inference time, which mismatching tokens can be safely accepted without compromising the final answer quality. We evaluate AutoJudge with multiple draft/target model pairs on mathematical reasoning and programming benchmarks, achieving significant speedups at the cost of a minor accuracy reduction. Notably, on GSM8K with the Llama 3.1 70B target model, our approach achieves up to 2 speedup over speculative decoding at the cost of a 1% drop in accuracy. When applied to the LiveCodeBench benchmark, AutoJudge automatically detects programming-specific important tokens, accepting 25 tokens per speculation cycle at a 2% drop in Pass@1. Our approach requires no human annotation and is easy to integrate with modern LLM inference frameworks.

Building robust and general reasoning ability is a central goal in the development of large language models (LLMs). Recent efforts increasingly turn to code as a rich training source, given its inherent logical structure and diverse reasoning paradigms--such as divide-and-conquer, topological ordering, and enumeration. However, reasoning in code is often expressed implicitly and entangled with syntactic or implementation noise, making direct training on raw code suboptimal. To address this, we introduce TraceMind, a large-scale corpus of 2.6 million samples that transforms code execution into explicit, step-by-step chain-of-thought style rationales, which we call Chain of Execution (CoE). The corpus spans domains including mathematics, classical algorithms and algorithmic competition, and is enriched with variable-tracing questions and code rewritings to enhance logical granularity and code diversity. We evaluate Tracepile using three training setups--continue-pretraining, instruction tuning after pretraining, and two-stage finetuning. Experiments across four base models (LLaMA 3, LLaMA 3.1, Qwen-2.5, and Qwen-2.5 Coder) and 20 benchmarks covering math, code, logic, and algorithms demonstrate consistent improvements. Notably, Tracepile boosts LLaMA3-8B by 9.2\% on average across nine math datasets and delivers clear gains on LiveCodeBench, CRUX, and Zebra Logic under two-stage finetuning.