arXiv subjects

Iddo Drori

Publications and source records attributed to Iddo Drori.

At least 19 recordsLinked to original sources

arXiv2026

Twin: Playing an Unknown Game with a Test-Time Digital Twin

We present a Test-time World-model Inference (Twin) system, in which a frontier coding agent writes an executable world model for completing continual learning tasks, such as ARC-AGI-3 games. Traditional approaches hand-engineer such models, one custom design per task. Each game hides its rules and goal, and our system constructs them from simulation and interaction alone. Its inductive prior over grid games is strong enough to recover the true transitions of the game and the goal on nearly all levels. Replay validation happens in a twin world model. The harness enforces that an action is not made until the program reproduces every previous observed game transition. Each mismatch between a world model prediction and the actual action result becomes a counterexample that is used to repair the world model. Twin clears 179 out of 183 levels (97.8%), and does so more efficiently than humans in 158 out of 179 levels (88.3%). The system infers the goal before any reward on 156 of the levels it clears (87.2%), and in the remaining levels automatically discovers the goal by search. The benchmark scores completion and action efficiency, between 0 and 100, against humans playing each game for the first time. Played directly, the base model scores only 7.8%; an off-the-shelf harness increases it to 61.1%, whereas our twin world model increases the same base model to 93.3%, clearing 23 out of 25 games. Building a usable world model is simpler than anticipated, whereas the harder problem is inferring the right goal.

cs.AI↗

arXiv2025

Diverse Inference and Verification for Advanced Reasoning

Reasoning LLMs such as OpenAI o1, o3 and DeepSeek R1 have made significant progress in mathematics and coding, yet find challenging advanced tasks such as International Mathematical Olympiad (IMO) combinatorics problems, Abstraction and Reasoning Corpus (ARC) puzzles, and Humanity's Last Exam (HLE) questions. We use a diverse inference approach that combines multiple models and methods at test time. We find that verifying mathematics and code problems, and rejection sampling on other problems is simple and effective. We automatically verify correctness of solutions to IMO problems by Lean, and ARC puzzles by code, and find that best-of-N effectively answers HLE questions. Our approach increases answer accuracy on IMO combinatorics problems from 33.3% to 77.8%, accuracy on HLE questions from 8% to 37%, and solves 80% of ARC puzzles that 948 humans could not and 26.5% of ARC puzzles that o3 high compute does not. Test-time simulations, reinforcement learning, and meta-learning with inference feedback improve generalization by adapting agent graph representations and varying prompts, code, and datasets. Our approach is reliable, robust, and scalable, and in the spirit of reproducible research, we will make it publicly available upon publication.

cs.AI↗

arXiv2024

AI-Driven Review Systems: Evaluating LLMs in Scalable and Bias-Aware Academic Reviews

Automatic reviewing helps handle a large volume of papers, provides early feedback and quality control, reduces bias, and allows the analysis of trends. We evaluate the alignment of automatic paper reviews with human reviews using an arena of human preferences by pairwise comparisons. Gathering human preference may be time-consuming; therefore, we also use an LLM to automatically evaluate reviews to increase sample efficiency while reducing bias. In addition to evaluating human and LLM preferences among LLM reviews, we fine-tune an LLM to predict human preferences, predicting which reviews humans will prefer in a head-to-head battle between LLMs. We artificially introduce errors into papers and analyze the LLM's responses to identify limitations, use adaptive review questions, meta prompting, role-playing, integrate visual and textual analysis, use venue-specific reviewing materials, and predict human preferences, improving upon the limitations of the traditional review processes. We make the reviews of publicly available arXiv and open-access Nature journal papers available online, along with a free service which helps authors review and revise their research papers and improve their quality. This work develops proof-of-concept LLM reviewing systems that quickly deliver consistent, high-quality reviews and evaluate their quality. We mitigate the risks of misuse, inflated review scores, overconfident ratings, and skewed score distributions by augmenting the LLM with multiple documents, including the review form, reviewer guide, code of ethics and conduct, area chair guidelines, and previous year statistics, by finding which errors and shortcomings of the paper may be detected by automated reviews, and evaluating pairwise reviewer preferences. This work identifies and addresses the limitations of using LLMs as reviewers and evaluators and enhances the quality of the reviewing process.

cs.AI↗

arXiv2023

Exploring the MIT Mathematics and EECS Curriculum Using Large Language Models

We curate a comprehensive dataset of 4,550 questions and solutions from problem sets, midterm exams, and final exams across all MIT Mathematics and Electrical Engineering and Computer Science (EECS) courses required for obtaining a degree. We evaluate the ability of large language models to fulfill the graduation requirements for any MIT major in Mathematics and EECS. Our results demonstrate that GPT-3.5 successfully solves a third of the entire MIT curriculum, while GPT-4, with prompt engineering, achieves a perfect solve rate on a test set excluding questions based on images. We fine-tune an open-source large language model on this dataset. We employ GPT-4 to automatically grade model responses, providing a detailed performance breakdown by course, question, and answer type. By embedding questions in a low-dimensional space, we explore the relationships between questions, topics, and classes and discover which questions and classes are required for solving other questions and classes through few-shot learning. Our analysis offers valuable insights into course prerequisites and curriculum design, highlighting language models' potential for learning and improving Mathematics and EECS education.

cs.CL↗

arXiv2022

Human Evaluation of Text-to-Image Models on a Multi-Task Benchmark

We provide a new multi-task benchmark for evaluating text-to-image models. We perform a human evaluation comparing the most common open-source (Stable Diffusion) and commercial (DALL-E 2) models. Twenty computer science AI graduate students evaluated the two models, on three tasks, at three difficulty levels, across ten prompts each, providing 3,600 ratings. Text-to-image generation has seen rapid progress to the point that many recent models have demonstrated their ability to create realistic high-resolution images for various prompts. However, current text-to-image methods and the broader body of research in vision-language understanding still struggle with intricate text prompts that contain many objects with multiple attributes and relationships. We introduce a new text-to-image benchmark that contains a suite of thirty-two tasks over multiple applications that capture a model's ability to handle different features of a text prompt. For example, asking a model to generate a varying number of the same object to measure its ability to count or providing a text prompt with several objects that each have a different attribute to identify its ability to match objects and attributes correctly. Rather than subjectively evaluating text-to-image results on a set of prompts, our new multi-task benchmark consists of challenge tasks at three difficulty levels (easy, medium, and hard) and human ratings for each generated image.

cs.CV↗

arXiv2022

From Human Days to Machine Seconds: Automatically Answering and Generating Machine Learning Final Exams

A final exam in machine learning at a top institution such as MIT, Harvard, or Cornell typically takes faculty days to write, and students hours to solve. We demonstrate that large language models pass machine learning finals at a human level, on finals available online after the models were trained, and automatically generate new human-quality final exam questions in seconds. Previous work has developed program synthesis and few-shot learning methods to solve university-level problem set questions in mathematics and STEM courses. In this work, we develop and compare methods that solve final exams, which differ from problem sets in several ways: the questions are longer, have multiple parts, are more complicated, and span a broader set of topics. We curate a dataset and benchmark of questions from machine learning final exams available online and code for answering these questions and generating new questions. We show how to generate new questions from other questions and course notes. For reproducibility and future research on this final exam benchmark, we use automatic checkers for multiple-choice, numeric, and questions with expression answers. We perform ablation studies comparing zero-shot learning with few-shot learning and chain-of-thought prompting using GPT-3, OPT, Codex, and ChatGPT across machine learning topics and find that few-shot learning methods perform best. We highlight the transformative potential of language models to streamline the writing and solution of large-scale assessments, significantly reducing the workload from human days to mere machine seconds. Our results suggest that rather than banning large language models such as ChatGPT in class, instructors should teach students to harness them by asking students meta-questions about correctness, completeness, and originality of the responses generated, encouraging critical thinking in academic studies.

cs.LG↗

arXiv2021

A Neural Network Solves, Explains, and Generates University Math Problems by Program Synthesis and Few-Shot Learning at Human Level

We demonstrate that a neural network pre-trained on text and fine-tuned on code solves mathematics course problems, explains solutions, and generates new questions at a human level. We automatically synthesize programs using few-shot learning and OpenAI's Codex transformer and execute them to solve course problems at 81% automatic accuracy. We curate a new dataset of questions from MIT's largest mathematics courses (Single Variable and Multivariable Calculus, Differential Equations, Introduction to Probability and Statistics, Linear Algebra, and Mathematics for Computer Science) and Columbia University's Computational Linear Algebra. We solve questions from a MATH dataset (on Prealgebra, Algebra, Counting and Probability, Intermediate Algebra, Number Theory, and Precalculus), the latest benchmark of advanced mathematics problems designed to assess mathematical reasoning. We randomly sample questions and generate solutions with multiple modalities, including numbers, equations, and plots. The latest GPT-3 language model pre-trained on text automatically solves only 18.8% of these university questions using zero-shot learning and 30.8% using few-shot learning and the most recent chain of thought prompting. In contrast, program synthesis with few-shot learning using Codex fine-tuned on code generates programs that automatically solve 81% of these questions. Our approach improves the previous state-of-the-art automatic solution accuracy on the benchmark topics from 8.8% to 81.1%. We perform a survey to evaluate the quality and difficulty of generated questions. This work is the first to automatically solve university-level mathematics course questions at a human level and the first work to explain and generate university-level mathematics course questions at scale, a milestone for higher education.

cs.LG↗

arXiv2021

Solving Linear Algebra by Program Synthesis

We solve MIT's Linear Algebra 18.06 course and Columbia University's Computational Linear Algebra COMS3251 courses with perfect accuracy by interactive program synthesis. This surprisingly strong result is achieved by turning the course questions into programming tasks and then running the programs to produce the correct answers. We use OpenAI Codex with zero-shot learning, without providing any examples in the prompts, to synthesize code from questions. We quantify the difference between the original question text and the transformed question text that yields a correct answer. Since all COMS3251 questions are not available online the model is not overfitting. We go beyond just generating code for questions with numerical answers by interactively generating code that also results visually pleasing plots as output. Finally, we automatically generate new questions given a few sample questions which may be used as new course content. This work is a significant step forward in solving quantitative math problems and opens the door for solving many university level STEM courses by machine.

cs.LG↗

arXiv2021

Solving Probability and Statistics Problems by Program Synthesis

We solve university level probability and statistics questions by program synthesis using OpenAI's Codex, a Transformer trained on text and fine-tuned on code. We transform course problems from MIT's 18.05 Introduction to Probability and Statistics and Harvard's STAT110 Probability into programming tasks. We then execute the generated code to get a solution. Since these course questions are grounded in probability, we often aim to have Codex generate probabilistic programs that simulate a large number of probabilistic dependencies to compute its solution. Our approach requires prompt engineering to transform the question from its original form to an explicit, tractable form that results in a correct program and solution. To estimate the amount of work needed to translate an original question into its tractable form, we measure the similarity between original and transformed questions. Our work is the first to introduce a new dataset of university-level probability and statistics problems and solve these problems in a scalable fashion using the program synthesis capabilities of large language models.

cs.LG↗

arXiv2021

Tracking Blobs in the Turbulent Edge Plasma of a Tokamak Fusion Device

The analysis of turbulence in plasmas is fundamental in fusion research. Despite extensive progress in theoretical modeling in the past 15 years, we still lack a complete and consistent understanding of turbulence in magnetic confinement devices, such as tokamaks. Experimental studies are challenging due to the diverse processes that drive the high-speed dynamics of turbulent phenomena. This work presents a novel application of motion tracking to identify and track turbulent filaments in fusion plasmas, called blobs, in a high-frequency video obtained from Gas Puff Imaging diagnostics. We compare four baseline methods (RAFT, Mask R-CNN, GMA, and Flow Walk) trained on synthetic data and then test on synthetic and real-world data obtained from plasmas in the Tokamak `a Configuration Variable (TCV). The blob regime identified from an analysis of blob trajectories agrees with state-of-the-art conditional averaging methods for each of the baseline methods employed, giving confidence in the accuracy of these techniques. High entry barriers traditionally limit tokamak plasma research to a small community of researchers in the field. By making a dataset and benchmark publicly available, we hope to open the field to a broad community in science and engineering.

physics.plasm-ph↗

arXiv2021

AlphaD3M: Machine Learning Pipeline Synthesis

We introduce AlphaD3M, an automatic machine learning (AutoML) system based on meta reinforcement learning using sequence models with self play. AlphaD3M is based on edit operations performed over machine learning pipeline primitives providing explainability. We compare AlphaD3M with state-of-the-art AutoML systems: Autosklearn, Autostacker, and TPOT, on OpenML datasets. AlphaD3M achieves competitive performance while being an order of magnitude faster, reducing computation time from hours to minutes, and is explainable by design.

cs.LG↗

arXiv2021

Predicting Critical Biogeochemistry of the Southern Ocean for Climate Monitoring

The Biogeochemical-Argo (BGC-Argo) program is building a network of globally distributed, sensor-equipped robotic profiling floats, improving our understanding of the climate system and how it is changing. These floats, however, are limited in the number of variables measured. In this study, we train neural networks to predict silicate and phosphate values in the Southern Ocean from temperature, pressure, salinity, oxygen, nitrate, and location and apply these models to earth system model (ESM) and BGC-Argo data to expand the utility of this ocean observation network. We trained our neural networks on observations from the Global Ocean Ship-Based Hydrographic Investigations Program (GO-SHIP) and use dropout regularization to provide uncertainty bounds around our predicted values. Our neural network significantly improves upon linear regression but shows variable levels of uncertainty across the ranges of predicted variables. We explore the generalization of our estimators to test data outside our training distribution from both ESM and BGC-Argo data. Our use of out-of-distribution test data to examine shifts in biogeochemical parameters and calculate uncertainty bounds around estimates advance the state-of-the-art in oceanographic data and climate monitoring. We make our data and code publicly available.

cs.LG↗

arXiv2021

Predicting Atlantic Multidecadal Variability

Atlantic Multidecadal Variability (AMV) describes variations of North Atlantic sea surface temperature with a typical cycle of between 60 and 70 years. AMV strongly impacts local climate over North America and Europe, therefore prediction of AMV, especially the extreme values, is of great societal utility for understanding and responding to regional climate change. This work tests multiple machine learning models to improve the state of AMV prediction from maps of sea surface temperature, salinity, and sea level pressure in the North Atlantic region. We use data from the Community Earth System Model 1 Large Ensemble Project, a state-of-the-art climate model with 3,440 years of data. Our results demonstrate that all of the models we use outperform the traditional persistence forecast baseline. Predicting the AMV is important for identifying future extreme temperatures and precipitation, as well as hurricane activity, in Europe and North America up to 25 years in advance.

cs.LG↗

arXiv2021

Solving the Families In the Wild Kinship Verification Challenge by Program Synthesis

Kinship verification is the task of determining whether a parent-child, sibling, or grandparent-grandchild relationship exists between two people and is important in social media applications, forensic investigations, finding missing children, and reuniting families. We demonstrate high quality kinship verification by participating in the 2021 Recognizing Families in the Wild challenge which provides the largest publicly available dataset in the field. Our approach is among the top 3 winning entries in the competition. We ensemble models written by both human experts and a foundation model, OpenAI Codex, trained on text and code. We use Codex to generate model variants, and also demonstrate its ability to generate entire running programs for kinship verification tasks of specific relationships.

cs.CV↗

arXiv2021

Pedestrian Wind Factor Estimation in Complex Urban Environments

Urban planners and policy makers face the challenge of creating livable and enjoyable cities for larger populations in much denser urban conditions. While the urban microclimate holds a key role in defining the quality of urban spaces today and in the future, the integration of wind microclimate assessment in early urban design and planning processes remains a challenge due to the complexity and high computational expense of computational fluid dynamics (CFD) simulations. This work develops a data-driven workflow for real-time pedestrian wind comfort estimation in complex urban environments which may enable designers, policy makers and city residents to make informed decisions about mobility, health, and energy choices. We use a conditional generative adversarial network (cGAN) architecture to reduce the computational computation while maintaining high confidence levels and interpretability, adequate representation of urban complexity, and suitability for pedestrian comfort estimation. We demonstrate high quality wind field approximations while reducing computation time from days to seconds.

cs.LG↗

arXiv2021

Image2Lego: Customized LEGO Set Generation from Images

Although LEGO sets have entertained generations of children and adults, the challenge of designing customized builds matching the complexity of real-world or imagined scenes remains too great for the average enthusiast. In order to make this feat possible, we implement a system that generates a LEGO brick model from 2D images. We design a novel solution to this problem that uses an octree-structured autoencoder trained on 3D voxelized models to obtain a feasible latent representation for model reconstruction, and a separate network trained to predict this latent representation from 2D images. LEGO models are obtained by algorithmic conversion of the 3D voxelized model to bricks. We demonstrate first-of-its-kind conversion of photographs to 3D LEGO models. An octree architecture enables the flexibility to produce multiple resolutions to best fit a user's creative vision or design needs. In order to demonstrate the broad applicability of our system, we generate step-by-step building instructions and animations for LEGO models of objects and human faces. Finally, we test these automatically generated LEGO sets by constructing physical builds using real LEGO bricks.

cs.CV↗

arXiv2021

Solving Machine Learning Problems

Can a machine learn Machine Learning? This work trains a machine learning model to solve machine learning problems from a University undergraduate level course. We generate a new training set of questions and answers consisting of course exercises, homework, and quiz questions from MIT's 6.036 Introduction to Machine Learning course and train a machine learning model to answer these questions. Our system demonstrates an overall accuracy of 96% for open-response questions and 97% for multiple-choice questions, compared with MIT students' average of 93%, achieving grade A performance in the course, all in real-time. Questions cover all 12 topics taught in the course, excluding coding questions or questions with images. Topics include: (i) basic machine learning principles; (ii) perceptrons; (iii) feature extraction and selection; (iv) logistic regression; (v) regression; (vi) neural networks; (vii) advanced neural networks; (viii) convolutional neural networks; (ix) recurrent neural networks; (x) state machines and MDPs; (xi) reinforcement learning; and (xii) decision trees. Our system uses Transformer models within an encoder-decoder architecture with graph and tree representations. An important aspect of our approach is a data-augmentation scheme for generating new example problems. We also train a machine learning model to generate problem hints. Thus, our system automatically generates new questions across topics, answers both open-response questions and multiple-choice questions, classifies problems, and generates problem hints, pushing the envelope of AI for STEM education.

cs.LG↗

arXiv2021

Privileged Zero-Shot AutoML

This work improves the quality of automated machine learning (AutoML) systems by using dataset and function descriptions while significantly decreasing computation time from minutes to milliseconds by using a zero-shot approach. Given a new dataset and a well-defined machine learning task, humans begin by reading a description of the dataset and documentation for the algorithms to be used. This work is the first to use these textual descriptions, which we call privileged information, for AutoML. We use a pre-trained Transformer model to process the privileged text and demonstrate that using this information improves AutoML performance. Thus, our approach leverages the progress of unsupervised representation learning in natural language processing to provide a significant boost to AutoML. We demonstrate that using only textual descriptions of the data and functions achieves reasonable classification performance, and adding textual descriptions to data meta-features improves classification across tabular datasets. To achieve zero-shot AutoML we train a graph neural network with these description embeddings and the data meta-features. Each node represents a training dataset, which we use to predict the best machine learning pipeline for a new test dataset in a zero-shot fashion. Our zero-shot approach rapidly predicts a high-quality pipeline for a supervised learning task and dataset. In contrast, most AutoML systems require tens or hundreds of pipeline evaluations. We show that zero-shot AutoML reduces running and prediction times from minutes to milliseconds, consistently across datasets. By speeding up AutoML by orders of magnitude this work demonstrates real-time AutoML.

cs.LG↗