arXiv subjects

Chris Cummins

Publications and source records attributed to Chris Cummins.

At least 19 recordsLinked to original sources

arXiv2026

AXLE: A Cloud Infrastructure for Lean 4 Theorem Proving Utilities

We present AXLE (Axiom Lean Engine), a cloud service for Lean 4 proof manipulation, extraction, and verification. Recent progress in AI for mathematics -- reinforcement learning pipelines, agentic proving workflows, dataset curation -- demands Lean 4 tooling that scales to millions of requests while remaining correct and robust; existing infrastructure offers parallel compilation but not scalable proof verification, higher-level proof manipulation, multi-version support, or per-request isolation at the throughput modern AI workflows require. AXLE provides 14 Lean 4 metaprogramming tools spanning strict proof verification, declaration metadata extraction, semantic source manipulation, deterministic proof repair and simplification, and lemma extraction. The service runs as a multi-tenant cloud deployment with per-request isolation and concurrent support for multiple Lean 4 and Mathlib versions, accessible via a Python SDK, command-line interface, web UI, MCP server, and raw HTTP API. AXLE is publicly available and free to use at https://axle.axiommath.ai and via the axiom-axle PyPI package, with no local Lean 4 installation required. It has served over 500 million requests to date and is the underlying infrastructure for Axiom Math's proving efforts, including its 12/12 score on the 2025 Putnam competition.

cs.LO↗

arXiv2026

ABC implies that Ramanujan's tau function misses almost all primes

Lehmer conjectured that Ramanujan's tau-function never vanishes. In a related direction, a folklore conjecture asserts that infinitely many primes arise as absolute values of Ramanujan's tau-function. Recently, Xiong showed that these prime values form a subset of the primes with density at most $2/11$. Assuming the $abc$ Conjecture, we prove the stronger upper bound \[ S(X):=\#\{\ell\le X:\ \ell\ \text{prime and } |\tau(n)|=\ell \text{ for some } n\ge 1\} = O(X^{13/22}), \] which implies that Ramanujan's tau-function misses a density 1 subset of the primes. We give a heuristic suggesting that $S(X)$ should nevertheless be infinite, with predicted order of magnitude \[ S(X)\asymp \frac{C X^{\frac{1}{11}}}{(\log X)^2}. \] The main engine in this note was formalized and produced automatically in Lean/Mathlib by AxiomProver from a natural-language statement of the problem.

math.NT↗

arXiv2026

Almost all primes are partially regular

For odd primes $p$, we let $K_p:=\mathbb{Q}(\zeta_p)$ be the $p$th cyclotomic field and let $\omega$ denote its Teichmuller character. For $\alpha>1/2$, we say that an odd prime $p$ is partially regular if the eigenspaces of the $p$-Sylow subgroup of $\operatorname{Cl}(K_p)$ under the Galois action vanish for all characters $\omega^{p-2k}$ with \[ 2\le 2k \le \frac{\sqrt{p}}{(\log p)^{\alpha}}. \] Equivalently, $p\nmid \operatorname{num}(B_{2k})$ throughout this range. We prove that a density-one subset of primes is partially regular in this sense. By Leopoldt reflection, this yields a partial Vandiver Theorem: for a density-one set of primes $p$, the even eigenspaces $A_p(\omega^{2k})$ vanish for all even $2k$ satisfying the inequality above. This result has consequences for Kubota-Leopoldt $p$-adic $L$-functions, congruences between cusp forms and Eisenstein series, and $p$-torsion in algebraic $K$-groups. The theorem proving partial regularity for almost all $p$ is fully formalized in Lean/Mathlib and was produced automatically by AxiomProver from a natural-language statement of the conjecture.

math.NT↗

arXiv2026

Dead ends in square-free digit walks

We study "dead ends" in square-free digit walks: square-free integers $N$ such that, in base $b$, every one-digit extension $bN+d$ is non-square-free. In base $10$, the stochastic independence model of Miller et al. suggests that infinite square-free walks occur with probability near $1$, corresponding to an asymptotic dead-end density of $\approx 5.218\times 10^{-5}$. We prove that the true asymptotic dead-end density satisfies \[ c_{\mathrm{dead}} \approx 1.317\times 10^{-9}, \] roughly a factor of $\sim 4\times 10^4$ smaller than the prediction. For every base $b\geq 2$, we prove that dead-end densities exist and are given by a closed-form expression (as a finite alternating sum of Euler products). The argument is fully formalized in Lean/Mathlib, and was produced automatically by AxiomProver from a natural-language statement of the problem.

math.CO↗

arXiv2026

Fel's Conjecture on Syzygies of Numerical Semigroups

Let $S=\langle d_1,\dots,d_m\rangle$ be a numerical semigroup and $k[S]$ its semigroup ring. The Hilbert numerator of $k[S]$ determines normalized alternating syzygy power sums $K_p(S)$ encoding alternating power sums of syzygy degrees. Fel conjectured an explicit formula for $K_p(S)$, for all $p\ge 0$, in terms of the gap power sums $G_r(S)=\sum_{g\notin S} g^r$ and universal symmetric polynomials $T_n$ evaluated at the generator power sums $\sigma_k=\sum_i d_i^k$ (and $\delta_k=(\sigma_k-1)/2^k$). We prove Fel's conjecture via exponential generating functions and coefficient extraction, solating the universal identities for $T_n$ needed for the derivation. The argument is fully formalized in Lean/Mathlib, and was produced automatically by AxiomProver from a natural-language statement of the conjecture.

math.CO↗

arXiv2025

CWM: An Open-Weights LLM for Research on Code Generation with World Models

We release Code World Model (CWM), a 32-billion-parameter open-weights LLM, to advance research on code generation with world models. To improve code understanding beyond what can be learned from training on static code alone, we mid-train CWM on a large amount of observation-action trajectories from Python interpreter and agentic Docker environments, and perform extensive multi-task reasoning RL in verifiable coding, math, and multi-turn software engineering environments. With CWM, we provide a strong testbed for researchers to explore the opportunities world modeling affords for improving code generation with reasoning and planning in computational environments. We present first steps of how world models can benefit agentic coding, enable step-by-step simulation of Python code execution, and show early results of how reasoning can benefit from the latter. CWM is a dense, decoder-only LLM trained with a context size of up to 131k tokens. Independent of its world modeling capabilities, CWM offers strong performance on general coding and math tasks: it reaches pass@1 scores of 65.8% on SWE-bench Verified (with test-time scaling), 68.6% on LiveCodeBench, 96.6% on Math-500, and 76.0% on AIME 2024. To support further research on code world modeling, we release model checkpoints after mid-training, SFT, and RL.

cs.SE↗

arXiv2025

What I cannot execute, I do not understand: Training and Evaluating LLMs on Program Execution Traces

Code generation and understanding are critical capabilities for large language models (LLMs). Thus, most LLMs are pretrained and fine-tuned on code data. However, these datasets typically treat code as static strings and rarely exploit the dynamic information about their execution. Building upon previous work on trace modeling, we study Execution Tuning (E.T.), a training procedure in which we explicitly model real-world program execution traces without requiring manual test annotations. We train and evaluate models on different execution trace granularities (line and instruction-level) and strategies on the task of output prediction, obtaining around 80% accuracy on CruxEval and MBPP, and showing the advantages of dynamic scratchpads (i.e., self-contained intermediate computations updated by the model rather than accumulated as a history of past computations) on long executions (up to 14k steps). Finally, we discuss E.T.'s practical applications.

cs.LG↗

arXiv2024

Finding Missed Code Size Optimizations in Compilers using LLMs

Compilers are complex, and significant effort has been expended on testing them. Techniques such as random program generation and differential testing have proved highly effective and have uncovered thousands of bugs in production compilers. The majority of effort has been expended on validating that a compiler produces correct code for a given input, while less attention has been paid to ensuring that the compiler produces performant code. In this work we adapt differential testing to the task of identifying missed optimization opportunities in compilers. We develop a novel testing approach which combines large language models (LLMs) with a series of differential testing strategies and use them to find missing code size optimizations in C / C++ compilers. The advantage of our approach is its simplicity. We offload the complex task of generating random code to an off-the-shelf LLM, and use heuristics and analyses to identify anomalous compiler behavior. Our approach requires fewer than 150 lines of code to implement. This simplicity makes it extensible. By simply changing the target compiler and initial LLM prompt we port the approach from C / C++ to Rust and Swift, finding bugs in both. To date we have reported 24 confirmed bugs in production compilers, and conclude that LLM-assisted testing is a promising avenue for detecting optimization bugs in real world compilers.

cs.SE↗

arXiv2024

Don't Transform the Code, Code the Transforms: Towards Precise Code Rewriting using LLMs

Tools for rewriting, refactoring and optimizing code should be fast and correct. Large language models (LLMs), by their nature, possess neither of these qualities. Yet, there remains tremendous opportunity in using LLMs to improve code. We explore the use of LLMs not to transform code, but to code transforms. We propose a chain-of-thought approach to synthesizing code transformations from a small number of input/output code examples that incorporates execution and feedback. Unlike the direct rewrite approach, LLM-generated transformations are easy to inspect, debug, and validate. The logic of the rewrite is explicitly coded and easy to adapt. The compute required to run code transformations is minute compared to that of LLM rewriting. We test our approach on 16 Python code transformations and find that LLM- generated transforms are perfectly precise for 7 of them and less imprecise than direct LLM rewriting on the others. We hope to encourage further research to improving the precision of LLM code rewriting.

cs.LG↗

arXiv2024

Meta Large Language Model Compiler: Foundation Models of Compiler Optimization

Large Language Models (LLMs) have demonstrated remarkable capabilities across a variety of software engineering and coding tasks. However, their application in the domain of code and compiler optimization remains underexplored. Training LLMs is resource-intensive, requiring substantial GPU hours and extensive data collection, which can be prohibitive. To address this gap, we introduce Meta Large Language Model Compiler (LLM Compiler), a suite of robust, openly available, pre-trained models specifically designed for code optimization tasks. Built on the foundation of Code Llama, LLM Compiler enhances the understanding of compiler intermediate representations (IRs), assembly language, and optimization techniques. The model has been trained on a vast corpus of 546 billion tokens of LLVM-IR and assembly code and has undergone instruction fine-tuning to interpret compiler behavior. LLM Compiler is released under a bespoke commercial license to allow wide reuse and is available in two sizes: 7 billion and 13 billion parameters. We also present fine-tuned versions of the model, demonstrating its enhanced capabilities in optimizing code size and disassembling from x86_64 and ARM assembly back into LLVM-IR. These achieve 77% of the optimising potential of an autotuning search, and 45% disassembly round trip (14% exact match). This release aims to provide a scalable, cost-effective foundation for further research and development in compiler optimization by both academic researchers and industry practitioners.

cs.PL↗

arXiv2024

Compiler generated feedback for Large Language Models

We introduce a novel paradigm in compiler optimization powered by Large Language Models with compiler feedback to optimize the code size of LLVM assembly. The model takes unoptimized LLVM IR as input and produces optimized IR, the best optimization passes, and instruction counts of both unoptimized and optimized IRs. Then we compile the input with generated optimization passes and evaluate if the predicted instruction count is correct, generated IR is compilable, and corresponds to compiled code. We provide this feedback back to LLM and give it another chance to optimize code. This approach adds an extra 0.53% improvement over -Oz to the original model. Even though, adding more information with feedback seems intuitive, simple sampling techniques achieve much higher performance given 10 or more samples.

cs.PL↗

arXiv2024

Priority Sampling of Large Language Models for Compilers

Large language models show great potential in generating and optimizing code. Widely used sampling methods such as Nucleus Sampling increase the diversity of generation but often produce repeated samples for low temperatures and incoherent samples for high temperatures. Furthermore, the temperature coefficient has to be tuned for each task, limiting its usability. We present Priority Sampling, a simple and deterministic sampling technique that produces unique samples ordered by the model's confidence. Each new sample expands the unexpanded token with the highest probability in the augmented search tree. Additionally, Priority Sampling supports generation based on regular expression that provides a controllable and structured exploration process. Priority Sampling outperforms Nucleus Sampling for any number of samples, boosting the performance of the original model from 2.87% to 5% improvement over -Oz. Moreover, it outperforms the autotuner used for the generation of labels for the training of the original model in just 30 samples.

cs.LG↗

arXiv2023

LoopTune: Optimizing Tensor Computations with Reinforcement Learning

Advanced compiler technology is crucial for enabling machine learning applications to run on novel hardware, but traditional compilers fail to deliver performance, popular auto-tuners have long search times and expert-optimized libraries introduce unsustainable costs. To address this, we developed LoopTune, a deep reinforcement learning compiler that optimizes tensor computations in deep learning models for the CPU. LoopTune optimizes tensor traversal order while using the ultra-fast lightweight code generator LoopNest to perform hardware-specific optimizations. With a novel graph-based representation and action space, LoopTune speeds up LoopNest by 3.2x, generating an order of magnitude faster code than TVM, 2.8x faster than MetaSchedule, and 1.08x faster than AutoTVM, consistently performing at the level of the hand-tuned library Numpy. Moreover, LoopTune tunes code in order of seconds.

cs.LG↗

arXiv2023

Rewriting History: Repurposing Domain-Specific CGRAs

Coarse-grained reconfigurable arrays (CGRAs) are domain-specific devices promising both the flexibility of FPGAs and the performance of ASICs. However, with restricted domains comes a danger: designing chips that cannot accelerate enough current and future software to justify the hardware cost. We introduce FlexC, the first flexible CGRA compiler, which allows CGRAs to be adapted to operations they do not natively support. FlexC uses dataflow rewriting, replacing unsupported regions of code with equivalent operations that are supported by the CGRA. We use equality saturation, a technique enabling efficient exploration of a large space of rewrite rules, to effectively search through the program-space for supported programs. We applied FlexC to over 2,000 loop kernels, compiling to four different research CGRAs and 300 generated CGRAs and demonstrate a 2.2$\times$ increase in the number of loop kernels accelerated leading to 3$\times$ speedup compared to an Arm A5 CPU on kernels that would otherwise be unsupported by the accelerator.

cs.PL↗

arXiv2023

Large Language Models for Compiler Optimization

We explore the novel application of Large Language Models to code optimization. We present a 7B-parameter transformer model trained from scratch to optimize LLVM assembly for code size. The model takes as input unoptimized assembly and outputs a list of compiler options to best optimize the program. Crucially, during training, we ask the model to predict the instruction counts before and after optimization, and the optimized code itself. These auxiliary learning tasks significantly improve the optimization performance of the model and improve the model's depth of understanding. We evaluate on a large suite of test programs. Our approach achieves a 3.0% improvement in reducing instruction counts over the compiler, outperforming two state-of-the-art baselines that require thousands of compilations. Furthermore, the model shows surprisingly strong code reasoning abilities, generating compilable code 91% of the time and perfectly emulating the output of the compiler 70% of the time.

cs.PL↗

arXiv2023

SLaDe: A Portable Small Language Model Decompiler for Optimized Assembly

Decompilation is a well-studied area with numerous high-quality tools available. These are frequently used for security tasks and to port legacy code. However, they regularly generate difficult-to-read programs and require a large amount of engineering effort to support new programming languages and ISAs. Recent interest in neural approaches has produced portable tools that generate readable code. However, to-date such techniques are usually restricted to synthetic programs without optimization, and no models have evaluated their portability. Furthermore, while the code generated may be more readable, it is usually incorrect. This paper presents SLaDe, a Small Language model Decompiler based on a sequence-to-sequence transformer trained over real-world code. We develop a novel tokenizer and exploit no-dropout training to produce high-quality code. We utilize type-inference to generate programs that are more readable and accurate than standard analytic and recent neural approaches. Unlike standard approaches, SLaDe can infer out-of-context types and unlike neural approaches, it generates correct code. We evaluate SLaDe on over 4,000 functions from ExeBench on two ISAs and at two optimizations levels. SLaDe is up to 6 times more accurate than Ghidra, a state-of-the-art, industrial-strength decompiler and up to 4 times more accurate than the large language model ChatGPT and generates significantly more readable code than both.

cs.PL↗

arXiv2023

BenchDirect: A Directed Language Model for Compiler Benchmarks

The exponential increase of hardware-software complexity has made it impossible for compiler engineers to find the right optimization heuristics manually. Predictive models have been shown to find near optimal heuristics with little human effort but they are limited by a severe lack of diverse benchmarks to train on. Generative AI has been used by researchers to synthesize benchmarks into existing datasets. However, the synthetic programs are short, exceedingly simple and lacking diversity in their features. We develop BenchPress, the first ML compiler benchmark generator that can be directed within source code feature representations. BenchPress synthesizes executable functions by infilling code that conditions on the program's left and right context. BenchPress uses active learning to introduce new benchmarks with unseen features into the dataset of Grewe's et al. CPU vs GPU heuristic, improving its acquired performance by 50%. BenchPress targets features that has been impossible for other synthesizers to reach. In 3 feature spaces, we outperform human-written code from GitHub, CLgen, CLSmith and the SRCIROR mutator in targeting the features of Rodinia benchmarks. BenchPress steers generation with beam search over a feature-agnostic language model. We improve this with BenchDirect which utilizes a directed LM that infills programs by jointly observing source code context and the compiler features that are targeted. BenchDirect achieves up to 36% better accuracy in targeting the features of Rodinia benchmarks, it is 1.8x more likely to give an exact match and it speeds up execution time by up to 72% compared to BenchPress. Both our models produce code that is difficult to distinguish from human-written code. We conduct a Turing test which shows our models' synthetic benchmarks are labelled as 'human-written' as often as human-written code from GitHub.

cs.LG↗

arXiv2023

Learning Compiler Pass Orders using Coreset and Normalized Value Prediction

Finding the optimal pass sequence of compilation can lead to a significant reduction in program size and/or improvement in program efficiency. Prior works on compilation pass ordering have two major drawbacks. They either require an excessive budget (in terms of compilation steps) at compile time or fail to generalize to unseen programs. In this paper, for code-size reduction tasks, we propose a novel pipeline to find program-dependent pass sequences within 45 compilation calls. It first identifies a coreset of 50 pass sequences via greedy optimization of a submodular function, and then learns a policy with Graph Neural Network (GNN) to pick the optimal sequence by predicting the normalized values of the pass sequences in the coreset. Despite its simplicity, our pipeline outperforms the default -Oz flag by an average of 4.7% over a large collection (4683) of unseen code repositories from diverse domains across 14 datasets. In comparison, previous approaches like reinforcement learning on the raw pass sequence space may take days to train due to sparse reward, and may not generalize well in held-out ones from different domains. Our results demonstrate that existing human-designed compiler flags can be improved with a simple yet effective technique that transforms the raw action space into a small one with denser rewards.

cs.PL↗