Minyang Tian

Minyang Tian, Luyu Gao, Shizhuo Dylan Zhang et al. · princeton, uw

Since language models (LMs) now outperform average humans on many challenging tasks, it has become increasingly difficult to develop challenging, high-quality, and realistic evaluations. We address this issue by examining LMs' capabilities to generate code for solving real scientific research problems. Incorporating input from scientists and AI researchers in 16 diverse natural science sub-fields, including mathematics, physics, chemistry, biology, and materials science, we created a scientist-curated coding benchmark, SciCode. The problems in SciCode naturally factorize into multiple subproblems, each involving knowledge recall, reasoning, and code synthesis. In total, SciCode contains 338 subproblems decomposed from 80 challenging main problems. It offers optional descriptions specifying useful scientific background information and scientist-annotated gold-standard solutions and test cases for evaluation. Claude3.5-Sonnet, the best-performing model among those tested, can solve only 4.6% of the problems in the most realistic setting. We believe that SciCode demonstrates both contemporary LMs' progress towards becoming helpful scientific assistants and sheds light on the development and evaluation of scientific AI in the future.

Minyang Tian, E. A. Huerta, Huihuo Zheng et al.

We present a new class of AI models for the detection of quasi-circular, spinning, non-precessing binary black hole mergers whose waveforms include the higher order gravitational wave modes $(l, |m|)=\{(2, 2), (2, 1), (3, 3), (3, 2), (4, 4)\}$, and mode mixing effects in the $l = 3, |m| = 2$ harmonics. These AI models combine hybrid dilated convolution neural networks to accurately model both short- and long-range temporal sequential information of gravitational waves; and graph neural networks to capture spatial correlations among gravitational wave observatories to consistently describe and identify the presence of a signal in a three detector network encompassing the Advanced LIGO and Virgo detectors. We first trained these spatiotemporal-graph AI models using synthetic noise, using 1.2 million modeled waveforms to densely sample this signal manifold, within 1.7 hours using 256 A100 GPUs in the Polaris supercomputer at the ALCF. Our distributed training approach had optimal performance, and strong scaling up to 512 A100 GPUs. With these AI ensembles we processed data from a three detector network, and found that an ensemble of 4 AI models achieves state-of-the-art performance for signal detection, and reports two misclassifications for every decade of searched data. We distributed AI inference over 128 GPUs in the Polaris supercomputer and 128 nodes in the Theta supercomputer, and completed the processing of a decade of gravitational wave data from a three detector network within 3.5 hours. Finally, we fine-tuned these AI ensembles to process the entire month of February 2020, which is part of the O3b LIGO/Virgo observation run, and found 6 gravitational waves, concurrently identified in Advanced LIGO and Advanced Virgo data, and zero false positives. This analysis was completed in one hour using one A100 GPU.

Akari Asai, Jacqueline He, Rulin Shao et al. · allen-ai

Scientific progress depends on researchers' ability to synthesize the growing body of literature. Can large language models (LMs) assist scientists in this task? We introduce OpenScholar, a specialized retrieval-augmented LM that answers scientific queries by identifying relevant passages from 45 million open-access papers and synthesizing citation-backed responses. To evaluate OpenScholar, we develop ScholarQABench, the first large-scale multi-domain benchmark for literature search, comprising 2,967 expert-written queries and 208 long-form answers across computer science, physics, neuroscience, and biomedicine. On ScholarQABench, OpenScholar-8B outperforms GPT-4o by 5% and PaperQA2 by 7% in correctness, despite being a smaller, open model. While GPT4o hallucinates citations 78 to 90% of the time, OpenScholar achieves citation accuracy on par with human experts. OpenScholar's datastore, retriever, and self-feedback inference loop also improves off-the-shelf LMs: for instance, OpenScholar-GPT4o improves GPT-4o's correctness by 12%. In human evaluations, experts preferred OpenScholar-8B and OpenScholar-GPT4o responses over expert-written ones 51% and 70% of the time, respectively, compared to GPT4o's 32%. We open-source all of our code, models, datastore, data and a public demo.

Yufeng Du, Phillip Harris, Minyang Tian et al.

We identify intrinsic limitations of Rotary Positional Embeddings (RoPE) in Transformer-based long-context language models. Our theoretical analysis abstracts away from the specific content of the context and depends only on its length. We prove that as context length increases, RoPE-based attention becomes unpredictable and loses two properties that are central to its effectiveness. First, it loses its locality bias: RoPE is no more likely to favor nearer positions than substantially farther ones. Second, it loses consistency in token relevance: a key vector that receives a higher attention score than an alternative at one position may receive a lower score at another. In both cases, the probability of failure approaches 0.5, no better than random guessing. We further prove that the attention score can remain unchanged when a key token is moved to a different position, or even replaced by a different token, indicating a failure to distinguish positions or tokens. Adjusting the RoPE base trades off distinguishing positions against distinguishing tokens but cannot preserve both at the same time. Increasing the RoPE base hyperparameter, a common practice in today's long-context models, helps distinguish different tokens, but inevitably sacrifices the ability to distinguish positions. Our empirical analysis shows that multi-head, multi-layer architectures are insufficient to overcome these limitations. Our findings suggest that fundamentally new mechanisms for encoding position and token order may be needed in future Transformer long-context language models.

Minyang Tian, E. A. Huerta, Huihuo Zheng

We introduce spatiotemporal-graph models that concurrently process data from the twin advanced LIGO detectors and the advanced Virgo detector. We trained these AI classifiers with 2.4 million IMRPhenomXPHM waveforms that describe quasi-circular, spinning, non-precessing binary black hole mergers with component masses $m_{\{1,2\}}\in[3M_\odot, 50 M_\odot]$, and individual spins $s^z_{\{1,2\}}\in[-0.9, 0.9]$; and which include the $(\ell, |m|) = \{(2, 2), (2, 1), (3, 3), (3, 2), (4, 4)\}$ modes, and mode mixing effects in the $\ell = 3, |m| = 2$ harmonics. We trained these AI classifiers within 22 hours using distributed training over 96 NVIDIA V100 GPUs in the Summit supercomputer. We then used transfer learning to create AI predictors that estimate the total mass of potential binary black holes identified by all AI classifiers in the ensemble. We used this ensemble, 3 classifiers for signal detection and 2 total mass predictors, to process a year-long test set in which we injected 300,000 signals. This year-long test set was processed within 5.19 minutes using 1024 NVIDIA A100 GPUs in the Polaris supercomputer (for AI inference) and 128 CPU nodes in the ThetaKNL supercomputer (for post-processing of noise triggers), housed at the Argonne Leadership Computing Facility. These studies indicate that our AI ensemble provides state-of-the-art signal detection accuracy, and reports 2 misclassifications for every year of searched data. This is the first AI ensemble designed to search for and find higher order gravitational wave mode signals.

Yufeng Du, Minyang Tian, Srikanth Ronanki et al.

Large language models (LLMs) often fail to scale their performance on long-context tasks performance in line with the context lengths they support. This gap is commonly attributed to retrieval failures -- the models' inability to identify relevant information in the long inputs. Accordingly, recent efforts often focus on evaluating and improving LLMs' retrieval performance: if retrieval is perfect, a model should, in principle, perform just as well on a long input as it does on a short one -- or should it? This paper presents findings that the answer to this question may be negative. Our systematic experiments across 5 open- and closed-source LLMs on math, question answering, and coding tasks reveal that, even when models can perfectly retrieve all relevant information, their performance still degrades substantially (13.9%--85%) as input length increases but remains well within the models' claimed lengths. This failure occurs even when the irrelevant tokens are replaced with minimally distracting whitespace, and, more surprisingly, when they are all masked and the models are forced to attend only to the relevant tokens. A similar performance drop is observed when all relevant evidence is placed immediately before the question. Our findings reveal a previously-unrealized limitation: the sheer length of the input alone can hurt LLM performance, independent of retrieval quality and without any distraction. They motivate our simple, model-agnostic mitigation strategy that transforms a long-context task into a short-context one by prompting the model to recite the retrieved evidence before attempting to solve the problem. On RULER, we observe a consistent improvement of GPT-4o up to 4% on an already strong baseline.

Franck Cappello, Sandeep Madireddy, Robert Underwood et al.

Recent advancements have positioned AI, and particularly Large Language Models (LLMs), as transformative tools for scientific research, capable of addressing complex tasks that require reasoning, problem-solving, and decision-making. Their exceptional capabilities suggest their potential as scientific research assistants but also highlight the need for holistic, rigorous, and domain-specific evaluation to assess effectiveness in real-world scientific applications. This paper describes a multifaceted methodology for Evaluating AI models as scientific Research Assistants (EAIRA) developed at Argonne National Laboratory. This methodology incorporates four primary classes of evaluations. 1) Multiple Choice Questions to assess factual recall; 2) Open Response to evaluate advanced reasoning and problem-solving skills; 3) Lab-Style Experiments involving detailed analysis of capabilities as research assistants in controlled environments; and 4) Field-Style Experiments to capture researcher-LLM interactions at scale in a wide range of scientific domains and applications. These complementary methods enable a comprehensive analysis of LLM strengths and weaknesses with respect to their scientific knowledge, reasoning abilities, and adaptability. Recognizing the rapid pace of LLM advancements, we designed the methodology to evolve and adapt so as to ensure its continued relevance and applicability. This paper describes the methodology state at the end of February 2025. Although developed within a subset of scientific domains, the methodology is designed to be generalizable to a wide range of scientific domains.

Minhui Zhu, Minyang Tian, Xiaocheng Yang et al.

While large language models (LLMs) with reasoning capabilities are progressing rapidly on high-school math competitions and coding, can they reason effectively through complex, open-ended challenges found in frontier physics research? And crucially, what kinds of reasoning tasks do physicists want LLMs to assist with? To address these questions, we present the CritPt (Complex Research using Integrated Thinking - Physics Test, pronounced "critical point"), the first benchmark designed to test LLMs on unpublished, research-level reasoning tasks that broadly covers modern physics research areas, including condensed matter, quantum physics, atomic, molecular & optical physics, astrophysics, high energy physics, mathematical physics, statistical physics, nuclear physics, nonlinear dynamics, fluid dynamics and biophysics. CritPt consists of 71 composite research challenges designed to simulate full-scale research projects at the entry level, which are also decomposed to 190 simpler checkpoint tasks for more fine-grained insights. All problems are newly created by 50+ active physics researchers based on their own research. Every problem is hand-curated to admit a guess-resistant and machine-verifiable answer and is evaluated by an automated grading pipeline heavily customized for advanced physics-specific output formats. We find that while current state-of-the-art LLMs show early promise on isolated checkpoints, they remain far from being able to reliably solve full research-scale challenges: the best average accuracy among base models is only 5.7%, achieved by GPT-5 (high), moderately rising to around 10% when equipped with coding tools. Through the realistic yet standardized evaluation offered by CritPt, we highlight a large disconnect between current model capabilities and realistic physics research demands, offering a foundation to guide the development of scientifically grounded AI tools.

Pranshu Chaturvedi, Asad Khan, Minyang Tian et al.

We introduce an ensemble of artificial intelligence models for gravitational wave detection that we trained in the Summit supercomputer using 32 nodes, equivalent to 192 NVIDIA V100 GPUs, within 2 hours. Once fully trained, we optimized these models for accelerated inference using NVIDIA TensorRT. We deployed our inference-optimized AI ensemble in the ThetaGPU supercomputer at Argonne Leadership Computer Facility to conduct distributed inference. Using the entire ThetaGPU supercomputer, consisting of 20 nodes each of which has 8 NVIDIA A100 Tensor Core GPUs and 2 AMD Rome CPUs, our NVIDIA TensorRT-optimized AI ensemble processed an entire month of advanced LIGO data (including Hanford and Livingston data streams) within 50 seconds. Our inference-optimized AI ensemble retains the same sensitivity of traditional AI models, namely, it identifies all known binary black hole mergers previously identified in this advanced LIGO dataset and reports no misclassifications, while also providing a 3X inference speedup compared to traditional artificial intelligence models. We used time slides to quantify the performance of our AI ensemble to process up to 5 years worth of advanced LIGO data. In this synthetically enhanced dataset, our AI ensemble reports an average of one misclassification for every month of searched advanced LIGO data. We also present the receiver operating characteristic curve of our AI ensemble using this 5 year long advanced LIGO dataset. This approach provides the required tools to conduct accelerated, AI-driven gravitational wave detection at scale.

E. A. Huerta, Asad Khan, Xiaobo Huang et al.

The development of reusable artificial intelligence (AI) models for wider use and rigorous validation by the community promises to unlock new opportunities in multi-messenger astrophysics. Here we develop a workflow that connects the Data and Learning Hub for Science, a repository for publishing AI models, with the Hardware Accelerated Learning (HAL) cluster, using funcX as a universal distributed computing service. Using this workflow, an ensemble of four openly available AI models can be run on HAL to process an entire month's worth (August 2017) of advanced Laser Interferometer Gravitational-Wave Observatory data in just seven minutes, identifying all four all four binary black hole mergers previously identified in this dataset and reporting no misclassifications. This approach combines advances in AI, distributed computing, and scientific data infrastructure to open new pathways to conduct reproducible, accelerated, data-driven discovery.