arXiv subjects

Alexander Novikov

Publications and source records attributed to Alexander Novikov.

At least 19 recordsLinked to original sources

arXiv2026

ArchAgent v2: A Case Study with the Data Prefetching Championship

Agentic artificial intelligence has shown great promise in automating algorithm design, but scaling similar techniques to computer microarchitecture discovery remains challenging due to vast search spaces, strict hardware budgets, and long simulation times. In this work, we present ArchAgent v2, a framework which scales automated microarchitecture search to multi-level data prefetching. While the original ArchAgent successfully discovered single-level cache replacement policies in competition settings, it does not scale to multi-level prefetching where the design space and degrees of freedom are larger. To overcome this, we introduce two new additions to ArchAgent: a cascaded evolutionary search that subdivides the design space by sequentially evolving and freezing prefetchers at individual cache levels, and a hardware-realizability feedback loop that embeds real-time size-estimation directly into the evolution process. Evaluated under identical rules of the 4th Data Prefetching Championship (DPC4), ArchAgent v2 automatically designs a three-level prefetcher that outperforms the winning hand-designed solution, further demonstrating automated agentic discovery as a useful tool for computer architects. Our discovered policy achieves a 3.8\% geometric mean IPC speedup over the baseline overall and a 0.3\% improvement over the prior champion, BertiGO. On low-bandwidth single-core configurations, our policy yields a 4.6\% performance speedup compared to only 2.6\% for BertiGO. However, multi-core evolution still remains a significant challenge due to simulation latency impeding evolution speed. Finally, our profiling of an ArchAgent evolution of over 12,000 candidate designs provides key insights into how automated evolutionary agents explore and synthesize complex microarchitectural logic.

cs.AI↗

arXiv2026

Intentmaking and Sensemaking: Human Interaction with AI-Guided Mathematical Discovery

Artificial intelligence offers powerful new tools for scientific discovery, but the interaction paradigms required to effectively harness these systems remain underexplored. In this paper, we present findings from a formative user study with 11 expert mathematicians who used AlphaEvolve, an evolutionary coding agent, to tackle advanced problems in their fields of expertise. We identify and characterize a distinct workflow we term intentmaking, the iterative process of discovering, defining, and refining one's experimental goals through active system interaction. We frame this as a natural extension to sensemaking, the cognitive process of building an understanding of complex or novel data. We suggest that users enter a cycle of intentmaking (defining and updating their experiment) and sensemaking (interpreting the results) which repeats many times during the course of an investigation. Our documentation of these themes suggests an approach to designing AI tools for scientific discovery that goes beyond the existing question/answer model of many current systems, treating them as collaborative instruments rather than opaque black-box assistants.

cs.AI↗

arXiv2026

ArchAgent: Agentic AI-driven Computer Architecture Discovery

Agile hardware design flows are a critically needed force multiplier to meet the exploding demand for compute. Recently, agentic generative AI systems have demonstrated significant advances in algorithm design, improving code efficiency, and enabling discovery across scientific domains. Bridging these worlds, we present ArchAgent, an automated computer architecture discovery system built on AlphaEvolve. We show ArchAgent's ability to automatically design/implement state-of-the-art (SoTA) cache replacement policies (architecting new mechanisms/logic, not only changing parameters), broadly within the confines of an established cache replacement policy design competition. In two days without human intervention, ArchAgent generated a policy achieving a 5.3% IPC speedup improvement over the prior SoTA on public multi-core Google Workload Traces. On the heavily-explored single-core SPEC06 workloads, it generated a policy in just 18 days showing a 0.9% IPC speedup improvement over the existing SoTA (a similar "winning margin" as reported by the existing SoTA). ArchAgent achieved these gains 3-5x faster than prior human-developed SoTA policies. Agentic flows also enable "post-silicon hyperspecialization" where agents tune runtime-configurable parameters exposed in hardware policies to further align the policies with a specific workload (mix). Exploiting this, we demonstrate a 2.4% IPC speedup improvement over prior SoTA on SPEC06 workloads. Finally, we outline broader implications for computer architecture research in the era of agentic AI. For example, we demonstrate the phenomenon of "simulator escapes", where the agentic AI flow discovered and exploited a loophole in a popular microarchitectural simulator - a consequence of the fact that these research tools were designed for a (now past) world where they were exclusively operated by humans acting in good-faith.

cs.AI↗

arXiv2026

Infrared photonics for healthcare: A roadmap for proactive and predictive health management

The field of infrared (IR) photonics is currently undergoing remarkable progress, moving rapidly towards practical sensing applications demanded by medical therapy and diagnostics (theranostics). The Developments can be divided into three main categories: (i) novel devices and measurement concepts including advanced updates of classical approaches that push medical sensing into the spotlight; (ii) new demonstrations of photonic integrated circuit (PIC-)based IR devices enabling highly miniaturized sensors for point-of-care application as well as medical and wellness wearables; and (iii) technologically-mature IR demonstrators that enable first medical sensing and treatment applications. This roadmap paper provides a consolidated overview of this highly dynamic and interdisciplinary research field with a focus on the major roadblocks that limit the widespread adoption of IR photonics in large-scale medical diagnostics. Special attention is given to the ambivalence between the molecular-level spectroscopic interpretation and a broader health-state assessment, highlighting the need for a common framework. Additionally, the paper discusses the critical importance of unified measurement standards, calibration protocols, and medical certification processes to ensure the validity of experimental results, reproducibility, and clinical trust, particularly when novel experimental techniques and AI algorithms are involved. Perspectives from major past and current contributors to application-oriented IR photonics will be provided.

physics.app-ph↗

arXiv2026

Magellan: Autonomous Discovery of Novel Compiler Optimization Heuristics with AlphaEvolve

Modern compilers rely on hand-crafted heuristics to guide optimization passes. These human-designed rules often struggle to adapt to the complexity of modern software and hardware and lead to high maintenance burden. To address this challenge, we present Magellan, an agentic framework that evolves the compiler pass itself by synthesizing executable C++ decision logic. Magellan couples an LLM coding agent with evolutionary search and autotuning in a closed loop of generation, evaluation on user-provided macro-benchmarks, and refinement, producing compact heuristics that integrate directly into existing compilers. Across several production optimization tasks, Magellan discovers policies that match or surpass expert baselines. In LLVM function inlining, Magellan synthesizes new heuristics that outperform decades of manual engineering for both binary-size reduction and end-to-end performance. In register allocation, it learns a concise priority rule for live-range processing that matches intricate human-designed policies on a large-scale workload. We also report preliminary results on XLA problems, demonstrating portability beyond LLVM with reduced engineering effort.

cs.AI↗

arXiv2025

AlphaEvolve: A coding agent for scientific and algorithmic discovery

In this white paper, we present AlphaEvolve, an evolutionary coding agent that substantially enhances capabilities of state-of-the-art LLMs on highly challenging tasks such as tackling open scientific problems or optimizing critical pieces of computational infrastructure. AlphaEvolve orchestrates an autonomous pipeline of LLMs, whose task is to improve an algorithm by making direct changes to the code. Using an evolutionary approach, continuously receiving feedback from one or more evaluators, AlphaEvolve iteratively improves the algorithm, potentially leading to new scientific and practical discoveries. We demonstrate the broad applicability of this approach by applying it to a number of important computational problems. When applied to optimizing critical components of large-scale computational stacks at Google, AlphaEvolve developed a more efficient scheduling algorithm for data centers, found a functionally equivalent simplification in the circuit design of hardware accelerators, and accelerated the training of the LLM underpinning AlphaEvolve itself. Furthermore, AlphaEvolve discovered novel, provably correct algorithms that surpass state-of-the-art solutions on a spectrum of problems in mathematics and computer science, significantly expanding the scope of prior automated discovery methods (Romera-Paredes et al., 2023). Notably, AlphaEvolve developed a search algorithm that found a procedure to multiply two $4 \times 4$ complex-valued matrices using $48$ scalar multiplications; offering the first improvement, after 56 years, over Strassen's algorithm in this setting. We believe AlphaEvolve and coding agents like it can have a significant impact in improving solutions of problems across many areas of science and computation.

cs.AI↗

arXiv2025

Full exceptional collections on the isotropic Grassmannians

We prove that the Kuznetsov--Polishchuk exceptional collections on rational homogeneous spaces of the symplectic groups $\mathrm{Sp}(2n,\mathbb{C})$ are full and consist of vector bundles. To achieve this, we construct several classes of complexes, which we call generalized staircase complexes, symplectic staircase complexes and secondary staircase complexes -- each of which may be of independent interest.

math.AG↗

arXiv2025

Plans for a new array of radio antennas for the detection of air showers at the 433m surface-detector array of the Pierre Auger Observatory

We present the design and science case for a new array of radio antennas to be located at the Pierre Auger Observatory. Six stations of three SKALA antennas each will be deployed around a single water-Cherenkov surface detector triggering the radio readout. The planned antenna layout will allow for the detection of cosmic rays above a few tens of PeV and reach full efficiency for vertical air showers at several hundred PeV in primary energy. The array will thus be a pathfinder to demonstrate that fully-efficient radio detection in combination with the underground muon detectors already present at the location is possible. This will enable combined studies of the muon component and the depth of shower maximum, relevant for hadronic interaction models studies and more accurate determination of the cosmic-ray mass composition in the energy range of the Galactic-to-extragalactic transition.

astro-ph.HE↗

arXiv2024

Secondary staircase complexes on isotropic Grassmannians

We introduce a class of equivariant vector bundles on isotropic symplectic Grassmannians $\mathrm{IGr}(k,2n)$ defined as appropriate truncations of staircase complexes and show that these bundles can be assembled into a number of complexes quasi-isomorphic to the symplectic wedge powers of the symplectic bundle on $\mathrm{IGr}(k,2n)$. We are planning to use these secondary staircase complexes to study fullness of exceptional collections in the derived categories of isotropic Grassmannians and Lefschetz exceptional collections on $\mathrm{IGr}(3,2n)$.

math.AG↗

arXiv2024

Amplifying human performance in combinatorial competitive programming

Recent years have seen a significant surge in complex AI systems for competitive programming, capable of performing at admirable levels against human competitors. While steady progress has been made, the highest percentiles still remain out of reach for these methods on standard competition platforms such as Codeforces. Here we instead focus on combinatorial competitive programming, where the target is to find as-good-as-possible solutions to otherwise computationally intractable problems, over specific given inputs. We hypothesise that this scenario offers a unique testbed for human-AI synergy, as human programmers can write a backbone of a heuristic solution, after which AI can be used to optimise the scoring function used by the heuristic. We deploy our approach on previous iterations of Hash Code, a global team programming competition inspired by NP-hard software engineering problems at Google, and we leverage FunSearch to evolve our scoring functions. Our evolved solutions significantly improve the attained scores from their baseline, successfully breaking into the top percentile on all previous Hash Code online qualification rounds, and outperforming the top human teams on several. Our method is also performant on an optimisation problem that featured in a recent held-out AtCoder contest.

cs.LG↗

arXiv2024

Quantum Circuit Optimization with AlphaTensor

A key challenge in realizing fault-tolerant quantum computers is circuit optimization. Focusing on the most expensive gates in fault-tolerant quantum computation (namely, the T gates), we address the problem of T-count optimization, i.e., minimizing the number of T gates that are needed to implement a given circuit. To achieve this, we develop AlphaTensor-Quantum, a method based on deep reinforcement learning that exploits the relationship between optimizing T-count and tensor decomposition. Unlike existing methods for T-count optimization, AlphaTensor-Quantum can incorporate domain-specific knowledge about quantum computation and leverage gadgets, which significantly reduces the T-count of the optimized circuits. AlphaTensor-Quantum outperforms the existing methods for T-count optimization on a set of arithmetic benchmarks (even when compared without making use of gadgets). Remarkably, it discovers an efficient algorithm akin to Karatsuba's method for multiplication in finite fields. AlphaTensor-Quantum also finds the best human-designed solutions for relevant arithmetic computations used in Shor's algorithm and for quantum chemistry simulation, thus demonstrating it can save hundreds of hours of research by optimizing relevant quantum circuits in a fully automated way.

quant-ph↗

arXiv2022

TAROGE-M: Radio Antenna Array on Antarctic High Mountain for Detecting Near-Horizontal Ultra-High Energy Air Showers

TAROGE-M is a self-triggered radio antenna array atop the 2700 m high Mt. Melbourne in Antarctica, designed to detect impulsive geomagnetic emission from extensive air showers induced by ultra-high energy (UHE) particles beyond 0.1 EeV, including cosmic rays (CRs), Earth-skimming tau neutrinos, and particularly, the "ANITA anomalous events" (AAEs) from near and below the horizon, which origin remains uncertain and requires more experimental inputs for clarification. The detection concept of TAROGE-M takes advantage of a high altitude with synoptic view toward the horizon as an efficient signal collector, and the radio quietness as well as strong and near vertical geomagnetic field in Antarctica. This approach has a low energy threshold, high duty cycle, and is easy to extend for quickly enlarging statistics. Here we report experimental results from the first TAROGE-M station deployed in 2020, corresponding to $25.3$-days of livetime. The station consists of six receiving antennas operating at 180-450 MHz, and can reconstruct source directions with $\sim0.3^\circ$ angular resolution. To demonstrate its ability to detect UHE air showers, a search for CR signals in the data was conducted, resulting in seven identified events. These events have a mean reconstructed energy of $0.95_{-0.31}^{+0.46}$ EeV and zenith angles between $25^\circ-82^\circ$, with both distributions agreeing with simulations. The estimated CR flux is also consistent with results of other experiments. The TAROGE-M sensitivity to AAEs is approximated by the tau neutrino exposure with simulations, suggesting comparable sensitivity as ANITA's at $~1$ EeV energy with a few station-years of operation. These first results verified the station design and performance in a polar and high-altitude environment, and are promising for further discovery of tau neutrinos and AAEs after an extension in the near future.

astro-ph.HE↗

arXiv2022

A Generalist Agent

Inspired by progress in large-scale language modeling, we apply a similar approach towards building a single generalist agent beyond the realm of text outputs. The agent, which we refer to as Gato, works as a multi-modal, multi-task, multi-embodiment generalist policy. The same network with the same weights can play Atari, caption images, chat, stack blocks with a real robot arm and much more, deciding based on its context whether to output text, joint torques, button presses, or other tokens. In this report we describe the model and the data, and document the current capabilities of Gato.

cs.AI↗

arXiv2021

Benchmarks for Deep Off-Policy Evaluation

Off-policy evaluation (OPE) holds the promise of being able to leverage large, offline datasets for both evaluating and selecting complex policies for decision making. The ability to learn offline is particularly important in many real-world domains, such as in healthcare, recommender systems, or robotics, where online data collection is an expensive and potentially dangerous process. Being able to accurately evaluate and select high-performing policies without requiring online interaction could yield significant benefits in safety, time, and cost for these applications. While many OPE methods have been proposed in recent years, comparing results between papers is difficult because currently there is a lack of a comprehensive and unified benchmark, and measuring algorithmic progress has been challenging due to the lack of difficult evaluation tasks. In order to address this gap, we present a collection of policies that in conjunction with existing offline datasets can be used for benchmarking off-policy evaluation. Our tasks include a range of challenging high-dimensional continuous control problems, with wide selections of datasets and policies for performing policy selection. The goal of our benchmark is to provide a standardized measure of progress that is motivated from a set of principles designed to challenge and test the limits of existing OPE methods. We perform an evaluation of state-of-the-art algorithms and provide open-source access to our data and code to foster future research in this area.

cs.LG↗

arXiv2021

Automatic differentiation for Riemannian optimization on low-rank matrix and tensor-train manifolds

In scientific computing and machine learning applications, matrices and more general multidimensional arrays (tensors) can often be approximated with the help of low-rank decompositions. Since matrices and tensors of fixed rank form smooth Riemannian manifolds, one of the popular tools for finding low-rank approximations is to use Riemannian optimization. Nevertheless, efficient implementation of Riemannian gradients and Hessians, required in Riemannian optimization algorithms, can be a nontrivial task in practice. Moreover, in some cases, analytic formulas are not even available. In this paper, we build upon automatic differentiation and propose a method that, given an implementation of the function to be minimized, efficiently computes Riemannian gradients and matrix-by-vector products between an approximate Riemannian Hessian and a given vector.

math.OC↗

arXiv2020

Semi-supervised reward learning for offline reinforcement learning

In offline reinforcement learning (RL) agents are trained using a logged dataset. It appears to be the most natural route to attack real-life applications because in domains such as healthcare and robotics interactions with the environment are either expensive or unethical. Training agents usually requires reward functions, but unfortunately, rewards are seldom available in practice and their engineering is challenging and laborious. To overcome this, we investigate reward learning under the constraint of minimizing human reward annotations. We consider two types of supervision: timestep annotations and demonstrations. We propose semi-supervised learning algorithms that learn from limited annotations and incorporate unlabelled data. In our experiments with a simulated robotic arm, we greatly improve upon behavioural cloning and closely approach the performance achieved with ground truth rewards. We further investigate the relationship between the quality of the reward model and the final policies. We notice, for example, that the reward models do not need to be perfect to result in useful policies.

cs.LG↗

arXiv2020

Offline Learning from Demonstrations and Unlabeled Experience

Behavior cloning (BC) is often practical for robot learning because it allows a policy to be trained offline without rewards, by supervised learning on expert demonstrations. However, BC does not effectively leverage what we will refer to as unlabeled experience: data of mixed and unknown quality without reward annotations. This unlabeled data can be generated by a variety of sources such as human teleoperation, scripted policies and other agents on the same robot. Towards data-driven offline robot learning that can use this unlabeled experience, we introduce Offline Reinforced Imitation Learning (ORIL). ORIL first learns a reward function by contrasting observations from demonstrator and unlabeled trajectories, then annotates all data with the learned reward, and finally trains an agent via offline reinforcement learning. Across a diverse set of continuous control and simulated robotic manipulation tasks, we show that ORIL consistently outperforms comparable BC agents by effectively leveraging unlabeled experience.

cs.LG↗

arXiv2020

Sydr: Cutting Edge Dynamic Symbolic Execution

The security development lifecycle (SDL) is becoming an industry standard. Dynamic symbolic execution (DSE) has enormous amount of applications in computer security (fuzzing, vulnerability discovery, reverse-engineering, etc.). We propose several performance and accuracy improvements for dynamic symbolic execution. Skipping non-symbolic instructions allows to build a path predicate 1.2--3.5 times faster. Symbolic engine simplifies formulas during symbolic execution. Path predicate slicing eliminates irrelevant conjuncts from solver queries. We handle each jump table (switch statement) as multiple branches and describe the method for symbolic execution of multi-threaded programs. The proposed solutions were implemented in Sydr tool. Sydr performs inversion of branches in path predicate. Sydr combines DynamoRIO dynamic binary instrumentation tool with Triton symbolic engine. We evaluated Sydr features on 64-bit Linux executables.

cs.CR↗