Peer Revenue Now accepting positions

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

All of arXiv

Sunday 8 November

Quantum Physics

Bridging Gaps with Tensor Networks: From the Haldane Gap and Beyond

Proofs generated by large language models are often difficult to understand, generalize, or formalize directly. Here we verify, digest, and formalize OpenAI's computer-assisted proof of a uniform gap in the periodic spin-1 Heisenberg chain using tensor-network diagrams and Lean 4. Reading the partition-function network along space and imaginary time shows how finite-ring bounds propagate across doubled lengths and inverse temperatures to all longer even rings, sharpening the spin-1 bound from the same finite inputs by a factor of 19 to $\Delta_L > 0.0729$ for all even $L \geq 36$. We identify reflection positivity as the general condition behind the proof and certify uniform gaps in seven additional spin chains and ladders. More broadly, digesting machine-generated proofs into transparent physical mechanisms yields sharper bounds, broader theorems, and reusable formal verifications.
Read the full abstract

2611.11049 · 8 Nov · 5+33 pages (main text: 1 figure, 1 table; Supplemental Material: 6 tables, about 25 tensor-network diagrams). Companion to the KITP talk "Bridging Gaps with Tensor Networks" (Intermediate-Scale AI for Physics, Oct 20, 2026): https://online.kitp.ucsb.edu/online/physicsai-m26/lu/ . Code, certificates, Lean 4 formalization: https://github.com/LionSR/CertifiedGaps

Analysis of PDEs

Almost sharp global well-posedness and low-regularity blow-up for the two-dimensional periodic Zakharov system

We study the Zakharov system on an arbitrary fixed rectangular two-dimensional torus with initial data in $H^s\times L^2\times H^{-1}$. For the Schrödinger regularity $s$, Kishimoto proved local well-posedness for $1/2\leq s\leq1$ and global well-posedness for $9/14<s<1$ below the Euclidean cubic ground-state mass. Under the same mass condition, we prove global well-posedness for every $1/2<s<1$, with polynomial bounds for the Sobolev norms. We add quartic and quintic corrections to Kishimoto's resonantly modified energy. The main difficulty is to cancel the interactions obstructing the global iteration without introducing singular coefficients near resonance or losing control of the higher-order terms generated by differentiation. The coefficients exploit cancellations between successive nonlinear interactions. Away from resonance, they cancel the problematic terms; near resonance, algebraic rearrangements avoid division by small phases and preserve sufficient frequency decay. For every $s<1/2$, we also construct solutions with arbitrarily small initial data in $H^s\times L^2\times H^{-1}$ whose wave component blows up in finite time while the Schrödinger component remains bounded in $H^s$. We combine explicit approximate solutions supported on three Fourier modes with stability at $H^{1/2}\times L^2$. Choosing successively larger frequencies controls the nonlinear interactions among infinitely many approximate solutions and makes the accumulated errors summable, while their wave amplitudes produce blow-up. These results identify $s=1/2$ as the threshold between global well-posedness and the existence of such blow-up solutions, with the endpoint unresolved.
Read the full abstract

High Energy Physics (Phenomenology)

Quark- and hadronic-level CPT violation phenomenology of neutral-mesons at at $d=5$ of the SME

This article summarizes an investigation of the connection between the hadron-level scalar effective field theory treatment of Lorentz symmetry violation in neutral-meson oscillations and the corresponding quark level $d=5$ nonminimal SME approach. The correspondence between the quark-level $a^{(5)}$ term is identified to first order in CPT violation (CPTV) based on the expectation that C, P, T symmetries and the spontaneously broken chiral symmetry of QCD should hold for both formalism. These can be related to the phenomenological parameter describing possible CPTV in the meson oscillations. A general result facilitating searches for kaons is given.
Read the full abstract

Optimization and Control

Integer Programming Parametrized by Imbalanced Circuit Packing

We study the standard form Integer Programming problem (IP) parametrized by the imbalanced circuit packing number. This number is $0$ if and only if $A$ can be transformed to a totally unimodular matrix by row operations, in which case (IP) is solvable in polynomial time. We show that if the constraint matrix represents a graphic matroid, (IP) is fixed-parameter tractable when parametrized by the imbalanced circuit packing number of the matrix. The proof combines an Erdős-Pósa property for edge-colored non-monochromatic cycles with a reduction to mixed-integer programming using few integer variables. In contrast, by a reduction from the parity potential problem, we show NP-completeness of feasibility of (IP) when the constraint matrix represents a cographic matroid, even if kernel of the constraint matrix is $2$-regular and imbalanced circuits pairwise intersect.
Read the full abstract

Databases

Agents Propose, Code Verifies, Humans Decide: Verified Agentic Ingestion of Heterogeneous Scientific Tables into Relational Databases

Public scientific datasets rarely arrive analysis-ready. They arrive as messy folders of wide matrices, long tables, spreadsheets, and per-cohort files whose keys and encodings differ, and every reanalysis must reconnect them first. A coding agent can automate this, but its mistakes can look plausible, and without an audit trail, a mistake that corrupts data can go uncaught. We present an AI-assisted ingestion pipeline built on one rule: LLM agents propose, deterministic code verifies, and humans decide. Narrow agents see compact, citable evidence about the files, never the data, and return typed proposals. A verifier with 39 check types tests each proposal against that evidence in two loops, one for the schema's structure and one for its meaning, and a human reviewer approves the schema and the load. Code compiles PostgreSQL DDL, reshapes the data, quarantines rows it cannot place, records provenance, and reconciles every staged row against the database. On four public biomedical datasets, one held out from development, 9 of 20 first proposals failed a check and every failure was caught: the loops repaired eight, and a one-line reviewer edit fixed the ninth. All 45.1 million values loaded with zero reconciliation gap. Discovering a dataset's structure took 4-7 LLM calls and preparing its data at most one more, costing under $0.50 per dataset. In a separate test, each of nine planted data errors was quarantined with a reason or recorded in an audit log.
Read the full abstract

All authors Bharatha Rankothge, Suzanne Hodgkins, Virginia Rich, Michael Kirby

Quantum Physics

Exact sampling beyond Fock inputs and linear optics: photon-number superpositions, photonic matrix-product states, adaptive and non-linear circuits

Clifford and Clifford's algorithm samples single-photon boson sampling exactly, one output mode at a time, by adding the input photons in a uniformly random order. We extend this construction to other input states and to adaptive and non-linear circuits. It applies whenever non-negative weights on output prefixes and input sets satisfy one marginal identity, a set-chain lemma: the random-order sampler is then exact at one weight evaluation per candidate mode, and it gives the posterior over the hidden input set. We build four exact samplers on it. For sources in a superposition of vacuum and one photon, the weights are rectangular permanents on an offset input set, and the cost has the exponent of single photons under loss. For photonic matrix-product states of bond dimension $\chi$, a gap-by-gap Kraus contraction turns each weight into a short sum of squared amplitudes. For adaptive linear optics, the chain starts from the posterior given the adaptive counts and rejects chains that contradict them. For a single-mode non-linear phase the identity fails, and rejection from linear branches is exact with acceptance $1/\|c\|_1^2$. All four samplers agree with strong simulation on small instances, and cost models put adaptive sampling with up to three adaptive modes within $10\%$ of single photons and a non-linear site at $\varphi=\pi/2$ at about $4\times$.
Read the full abstract

Distributed Computing

GARDEN: Efficient GPU-Accelerated Billion-Scale GNN Training via Caching-Sampling-Batching Co-Design

Graph Neural Networks (GNNs) have shown impressive performance in diverse tasks such as social network analysis and recommendation systems. However, scaling their training to billion-edge graphs is hindered by the memory wall: GPU memory capacity falls short of the massive storage requirements for node features and topology. Sampling-based GNN training frameworks mitigate this by streaming subgraphs from CPU to GPU, but they incur prohibitive PCIe data transfers and non-coalesced memory accesses, resulting in bandwidth underutilization and low compute efficiency. We propose GARDEN, a GPU-accelerated sampling-based system that co-designs caching, sampling, and batching to eliminate these bottlenecks. GARDEN introduces three key innovations: (1) an Adaptive High-Fidelity Cache Allocator (AHFCA) that combines offline reuse profiling and online hotness tracking to dynamically cache features and adjacency lists within a fixed GPU memory budget, avoiding static allocation waste; (2) a Grouping-based Parallel Sampler that reorders stochastic neighborhood accesses(GPS) to maximize warp occupancy and hide latency on power-law graphs; and (3) a Super-batch Degree Classifier(SDC) that fuses mini-batches into degree-homogeneous groups, enabling cache-friendly coalesced transfers and high GPU utilization. Evaluations on real-world billion-scale graphs demonstrate that GARDEN achieves 4.54x, 1.27x, and 2.97x end-to-end training speedups over DGL-UVA, DUCATI, and FastGL, respectively, while maintaining model convergence and accuracy.
Read the full abstract

All authors Qingrui Zhu, Lei Liu, Yushu Cai, Kai Sheng, Zhiwen Chen, Xin He

Applied Physics

$>3\text{ W/mm}$ Output Power $\mathrm{Al}_{0.61}\mathrm{Ga}_{0.39}\mathrm{N}$ Channel HEMT with $24.4\%$ Associated PAE at $10\text{ GHz}$

In this work, an ultra-wide bandgap (UWBG) high-electron-mobility transistor (HEMT) was fabricated using an $\mathrm{Al}_{0.76}\mathrm{Ga}_{0.24}\mathrm{N}$ barrier and an $\mathrm{Al}_{0.61}\mathrm{Ga}_{0.39}\mathrm{N}$ channel with thick $\mathrm{SiO}_2$ passivation. The device demonstrated low current collapse at $40\text{ V}$ drain bias. At $10\text{ GHz}$, the device exhibited an output power of $3.07\text{ W/mm}$ with an associated power-added efficiency (PAE) of $24.4\%$. A peak PAE of $33.9\%$ was achieved at an associated output power of $2.14\text{ W/mm}$. This excellent large-signal performance highlights the strong potential of UWBG $\mathrm{AlGaN}$ channel devices for X-band applications.
Read the full abstract

All authors Meherab Hossain, Ruixin Bai, Swarnav Mukhopadhyay, Brahmani Challa, Kuan Chang Pan, Ryan Gilbert, Guangying Wang, Surjava Sanyal, Chuang Ye, Md Tahmidul Alam, Khush Gohel, Ahmad E. Islam, Andrew J. Green, Shubhra S. Pasayat, Chirag Gupta

Computer Vision

Virtual Task Prompting for Multi-Task Scene Understanding

Multi-task scene understanding models typically rely on modified vision encoders or complex decoders for cross-task interaction, which limits their flexibility and compatibility with off-the-shelf vision foundation models (VFMs). We introduce Virtual Task Prompting (VTP), which frames multi-task prompting as retrieving task-adaptive compositions from a shared virtual state, where both the state and routing process are optimized in an end-to-end manner. Concretely, VTP maintains an expanded set of virtual tasks alongside the backbone. At each layer, learnable routing matrices read task-specific compositions as input to interact with image tokens and write the processed results back, forming explicit pathways that capture the intricacies among tasks outside the encoder. Since VTP keeps the backbone clean, it integrates seamlessly with pretrained VFMs, advancing the state of the art on Pascal-Context and NYUD-v2. Stress test on Taskonomy with 13 heterogeneous tasks further shows VTP's robustness in extreme settings. Ablations and analysis indicate that VTP's gains stem from adaptive routing rather than increased capacity. The project page is available at https://jacky1128.github.io/VTP/.
Read the full abstract

Software Engineering

Advancements in Cloud-Native Systems: A Systematic Mapping Study of Reliability and Scalability across Cloud-Native, Edge-Native and AI-Enabled Systems

Cloud-native systems now extend to the network edge and host artificial-intelligence (AI) workloads, including AI agents, but evidence on their reliability and scalability is scattered and uneven. We report a PRISMA 2020 systematic mapping study with an evidence-graded synthesis. Five logged OpenAlex searches on 11 October 2026 covered cloud-native systems, edge AI, agentic AI for operations, reliability of agentic applications and physical AI; 189 records from an earlier exploratory search were added separately. After 301 duplicate removals and an automated peer-review venue rule, 4,611 records were screened against written criteria and a five-level relevance scale. The review includes 1,794 primary studies from 2017-2026 and 95 secondary studies, with every decision released in Dataset S1. Each primary study was assigned to a 13-category taxonomy and graded as simulation, prototype or testbed, production data, production deployment, or not determinable from the abstract. Of 1,003 studies with a determinable level, 66.2% used prototypes or testbeds, 18.0% simulations and 15.8% production data or deployments. A 93-study synthesis core, appraised from indexed abstracts, shows production deployments of cascading-failure prevention, trace- and graph-based root-cause analysis and predictive autoscaling, and production-trace evaluations of cold-start mitigation. Edge reliability and physical AI rely mainly on simulations and testbeds. AI-agent reliability uses tool-grounded diagnosis, explicit safety properties, trace-driven repair and workload-aware scheduling, while empirical studies show that agent frameworks remain fragile. Conclusions are limited to the documented corpus. The principal gaps are production evidence for edge-native systems, shared resilience benchmarks and dependable agentic operations.
Read the full abstract

Differential Geometry

Jacobi fields and Jacobi sections for semisprays on Lie algebroids: transverse variations and equivalences

We study the relationship between Jacobi sections and Jacobi fields associated with semisprays on Lie algebroids. First, we characterize compatible transverse variations in terms of Lie algebroid morphisms. This characterization leads to a variational description of Jacobi sections: a section along an integral curve of a semispray is a Jacobi section if and only if it arises as the infinitesimal variation of a compatible transverse variation of a dynamical variation satisfying a suitable condition on its image in the base. We also investigate the relationship between Jacobi sections and Jacobi fields through the anchor map. We prove that every Jacobi section projects to a Jacobi field, and establish a converse when the corresponding base variation is contained in a single leaf of the characteristic foliation.
Read the full abstract

Information Theory

Iterative Interference Cancellation for Near-Field XL-MIMO Networks with LoS Channels

In extra-large multiple-input multiple-output (XL-MIMO) systems, near-field users experience spherical wavefront propagation, offering higher spatial resolution than far-field plane-wave models, but complicating signal detection and interference cancellation. Under line-of-sight conditions, we propose a weighted-rate successive interference cancellation (WR-SIC) scheme whose ordering metric exploits the enhanced spatial degrees of freedom of near-field propagation to prioritize users with stronger channels. An iterative detection and decoding (IDD) scheme with LDPC coding is then developed. Simulations in hybrid near- and far-field deployments show that the proposed WR-SIC and IDD schemes improve bit-error and sum rates over existing schemes
Read the full abstract

Applied Physics

Demonstration of High fmax (> 80 GHz) in Recessed-Gate and Passivated $\mathrm{Al_{0.60}Ga_{0.40}N}$ Channel RF HEMTs

In this work, ultra-wide bandgap (UWBG) $\mathrm{Al_{0.60}Ga_{0.40}N}$ channel high-electron-mobility transistors (HEMTs) with an $\mathrm{Al_{0.81}Ga_{0.19}N}$ barrier and an $\mathrm{Al_{0.86}Ga_{0.14}N}$ back barrier were fabricated. The devices featured a recessed T-gate structure and $\mathrm{SiO_2}$ passivation. Scaled devices achieved a cutoff frequency ($f_{\mathrm{T}}$) of 35 GHz, a maximum oscillation frequency ($f_{\mathrm{max}}$) of 82 GHz, and an average breakdown field $> 3 \mathrm{MV/cm}$, while maintaining a reasonable current collapse of $\sim 20\%$ under a 40 V drain bias in pinch-off condition. Devices with larger dimensions demonstrated a high off-state breakdown voltage of 883 V and an $f_{\mathrm{T}}$ of 24 GHz, yielding a high Johnson's figure of merit ($\mathrm{JFOM}$) of 21.19 $\mathrm{THz \cdot V}$. Across all device geometries, a high $I_{\mathrm{ON}}/I_{\mathrm{OFF}}$ ratio exceeding $10^6$ was observed with drain and gate leakage currents strictly below $10^{-6} \mathrm{A/mm}$. This work demonstrates an optimized design for $\mathrm{AlGaN}$ channel RF HEMTs and their excellent potential for X- to Ka-band applications.
Read the full abstract

All authors Meherab Hossain, Ruixin Bai, Swarnav Mukhopadhyay, Brahmani Challa, Guangying Wang, Surjava Sanyal, Md Tahmidul Alam, Khush Gohel, Rafael A. Choudhury, Shubhra S. Pasayat, Chirag Gupta

Hardware Architecture

When Is High-Bandwidth Flash Cost-Effective for LLM Inference?

Can high-bandwidth flash (HBF) make LLM inference cheaper than adding GPUs? By storing weights closer to the GPU than SSD, HBF may reduce the number of GPUs needed for model capacity. However, this integration is only justified if it meets service targets and its cost is recovered. We simulate this integration using public component data across nine dense and mixture-of-experts (MoE) models, five workloads, two GPU platforms, and 1–100 GPUs. At fixed demand, HBF reduces the minimum GPU count to meet a service target in 98 of 468 comparisons; majority of them are paused and short sessions, 63 remain cost-effective at half a GPU price per HBF stack. For single short-chat users, HBF enables large MoEs, e.g. DeepSeek V3 and Kimi K2, to meet a relaxed service target on one RTX~4090 GPU, where conventional serving fails. At fixed GPU count, HBF improves throughput in 48% of matched comparisons below 16 GPUs and 13% at larger counts. Even at 100 RTX~4090-scale GPUs, cascaded HBF raises offline throughput by 16.4–26.5$\times$ for three large MoEs. We found savings and throughput gains arise in different scenarios.
Read the full abstract

Computation and Language

Room to Reconsider: Reply Length Decides Whether Pushback Makes Language Models Sycophantic or Self-Correcting

Language models often abandon a correct answer when a user pushes back without giving a reason, and this behavior is usually studied as sycophancy to be suppressed. The same pushback, however, also leads models to revisit and sometimes repair a wrong answer. We measure both effects on multiple-choice questions from MMLU and ARC-Challenge with four open models from two families (Qwen3 1.7B, 4B and 8B, and OLMo-2 7B), reading the final answer exactly from option logits after a free-form reply. The reply budget changes the verdict. When Qwen3-4B may answer the challenge in at most 96 tokens, pushback leaves accuracy unchanged or slightly lower; with 1,024 tokens, it corrects 43% of its wrong answers and pushback raises accuracy by 5.7 points. Whether pushback helps overall follows a simple break-even rule in the cave and correction rates. Inside the model, caving can be partly predicted before the challenge: hidden states add about 0.04-0.06 AUROC over turn-1 confidence in three of our four probing runs. A probe trained only on self-correction cases predicts caving beyond confidence when replies are truncated, but not when the model can finish its reasoning. Activation steering never reduced caving selectively: where it worked, it suppressed self-correction as well. We conclude that sycophancy evaluations should allow full replies and report self-correction alongside caving.
Read the full abstract

Machine Learning

Prune to Probe: Causal Decision-Stability Probes for Efficient Reasoning

Reasoning models often continue generating after producing an answer, yet stopping prematurely may interrupt useful verification or revision. In this work, we investigate whether targeted interventions in a model's internal computation can provide a reliable signal for when to stop reasoning. We introduce Prune-to-Probe, an inference-time controller that uses causal activation interventions to probe the stability of a model's current answer. Specifically, we identify late-MLP coordinates that contribute to post-answer continuation while excluding those important for answer completion. Once an answer is stated, Prune-to-Probe temporarily suppresses the selected coordinates and measures the resulting stopping response relative to matched random interventions. Crucially, these interventions serve only as probes: the model either returns its existing answer or resumes generation from an untouched cache. On MATH, Prune-to-Probe reduces reasoning-token forwards by 2.60%, preserving all naturally correct trajectories while repairing six errors. On MMLU-Pro, it achieves 1.145% pooled savings with no observed harm. Further analysis shows that the intervention captures decision stability rather than correctness (AUROC 0.502). These results demonstrate the potential of causal activation probes for efficient inference-time control without permanent model modification.
Read the full abstract

High Energy Physics (Phenomenology)

Quark-level CPT violation in neutral-meson systems

This work investigates CPT and Lorentz breakdown in neutral-meson interferometry within the nonminimal Standard-Model Extension. Employing methods of chiral perturbation theory for mass dimension d=5, a connection between the quark-level and the meson effective pseudoscalar Lagrangian is established. The resulting phenomenological signatures in neutral-meson oscillations with particular focus on the detector geometry are explored. This analysis permits the interpretation of such CPT tests within the d=5 quark sector of the Standard-Model Extension. Existing data are used to extract first constraints on two coefficients and to estimate achievable sensitivity gains over published bounds of up to 18 orders of magnitude.
Read the full abstract

Machine Learning

Finite-Sample and Communication-Efficient Networked Information Aggregation

Building on SODA 2026 work Kearns et al. (2026), we study learning in directed acyclic graphs in which each agent observes only a subset of the input coordinates and may use the predictions of its parents as additional features. In the finite-sample protocol, the message sent along each edge is the agent's prediction vector on a shared sample, so the communication cost is linear in the sample size $m$. In this paper we study this model under communication constraints. First, we revisit the finite-sample analysis of the full-message protocol. We identify a gap in the finite-sample proof of Kearns et al. (2026) and develop a different argument that establishes a finite-sample population-risk guarantee along paths satisfying deterministic $M$-coverage. Second, we introduce a shared-sketch protocol in which each agent sends a $k$-dimensional linear sketch of its prediction vector instead of the full $m$-dimensional message, where $k\ll m$. We analyze this protocol through a reduction to least squares on the compressed sample together with an oblivious subspace embedding argument. The resulting bound separates statistics from communication: the statistical term is still governed by the original sample size $m$, while the additional error due to compression is controlled by the sketch dimension $k$. For Gaussian sketches, we obtain an explicit high-probability population-risk bound, showing that row subsampling is not the only possible communication-statistics tradeoff for networked information aggregation.
Read the full abstract

All authors Mohammadhossein Bateni, Zahra Hadizadeh, MohammadTaghi Hajiaghayi, Mahdi JafariRaviz, Shayan Taherijam

Machine Learning

Hybrid Product Search Without the Copy: Lexical, Semantic and Filtered Retrieval Served Directly from a Data Lake

E-commerce catalogs increasingly live in lakehouse tables: Apache Iceberg metadata over Parquet files in object storage. Product search, however, still runs on copies of that catalog, in a lexical engine and a vector database, each with its own ingestion pipeline, schema, staleness window and bill. We show that search applications can be hosted directly on the data lake, with the lakehouse table as the single system of record. We extend Apache Pinot, a real-time OLAP engine that already serves Iceberg tables in place, with BM25 relevance scoring over its Lucene text index, corpus-wide term statistics so that scores are comparable across files, and reciprocal rank fusion in plain SQL; a real-time table fed by a stream can sit behind the same table, merged by the broker. On the Amazon ESCI benchmark (1.2M US products, 22,458 test queries), hybrid search served from the lake matches exact retrieval (nDCG@10 .361). The lake table is queryable 149 s after its files are registered, and a lake commit is searchable in 3.8 s, close to indexing directly into a search engine. On a brand-sorted lake of 277M products (0.8 TB), one 16-vCPU node answers brand-scoped vector, BM25 and client-fused hybrid searches in 6.5-9 ms (p50), sustains 1,100 vector and 635 BM25 searches per second at under 180 ms p99, and makes a commit searchable in 4.3 s. Search adds only its indexes: no copy of the catalog, no sync pipeline and no second schema, which cuts storage, rebuild time and the operational work of keeping copies consistent.
Read the full abstract

Computation and Language

Building and Auditing Chemical Datasets with LLMs: A General LLM-as-a-Judge Framework Applied to Homogeneous Catalytic PET Depolymerisation

Large language models can rapidly extract structured multivariate experimental data from the chemistry literature, yet verifying thousands of generated records without ground truth remains difficult. We present a general, citation-grounded extraction and verification methodology centered on an independent "LLM-as-a-judge", applied to the literature on homogeneous catalytic poly(ethylene terephthalate) (PET) depolymerisation. In this pipeline, an extractor LLM generates schema-enforced records from research papers with citations to document chunks tagged with universally unique identifiers (UUIDs); an independent LLM judge then reads the full paper alongside each record, flags unsupported fields, and proposes cited corrections. Prompt ablations show that requiring chunk-level citations not only provides provenance down to specific paragraphs and tables but also improves extraction accuracy. Before deploying the judge at scale, we benchmarked it against manual evaluation on extractions from a hand-curated reference set of 295 ionic-liquid-catalysed PET glycolysis experiments. The judge agreed with human evaluation on 85.7% of reviewed records and was significantly more effective than a heuristic string-matching baseline at catching errors across related experiments. Deployed across 3,942 extracted records from 561 papers, the framework accepted 67.5% of them as written. Its proposed corrections concentrated on quantitative fields, where the extractor often conflated related experiments. We release the extracted dataset with the complete LLM judge audit logs, as well as the Chemistry Data Extractor Toolkit, an open-source application implementing this end-to-end pipeline in a code-free, browser-based environment for document parsing, customizable schema and prompt design, model-agnostic extraction and auditing, interactive provenance-highlighted review, and reproducible dataset bundling.
Read the full abstract

All authors Muhammad bin Javaid, Karim K. Ben Hicham, Enkhnyam Battulga, Alexander Mitsos, Martin Grohe

Physics and Society

Evolutionary games on networks with negative edge-based environmental feedback

Environmental feedback can fundamentally alter the game payoff structure underlying the evolution of cooperation. How feedback-evolving games interact with population structure under non-weak selection remains incompletely understood. We study an evolutionary game on networks in which each edge has a binary environmental state that determines the payoff matrix used by the two connected individuals. By assumption, cooperation tends to improve the local environment, which in turn favors defection, thereby creating negative feedback between strategies coupled with the environmental condition. We analyze the model under strong selection using a mean-field approximation, a pair approximation that tracks both strategy pairs and edge environments, and agent-based simulations. In well-mixed populations, the feedback can generate bistability between full cooperation and full defection, as well as sustained oscillations arising through a Hopf bifurcation. Sparse networks (i.e., with low nodes' degrees) can eliminate bistability, suppress oscillations, and produce stable intermediate levels of cooperation. The pair approximation captures these effects and generally agrees more closely with simulations on sparse networks than the mean-field approximation. Our results show that edge-based environmental feedback and network structure, such as density of edges, interact to determine whether the fraction of cooperators stabilizes or oscillates over time.
Read the full abstract

High Energy Physics (Theory)

Kinematically Unrestricted Flat Space Scattering from AdS$_{d+1} \times$ S$^p$/CFT$_d$

We construct scattering states for scalar fields with generic kinematics from CFT operators in the flat space limit of AdS/CFT. The states can have momentum along the compact $S^p$ directions of $AdS_{d+1}\times S^p$, which is achieved by scaling to large R-symmetry charge and choosing a particular parameterization for the R-symmetry polarization. These states allow for the construction of an unrestricted S-matrix from the flat space limit of CFT correlation functions, which we check for various tree level interactions. We also describe a generalization of the Mellin space formalism which incorporates the compact momentum, and we discuss some possible implications for flat space holography.
Read the full abstract

Computation and Language

MoTo: Mixture of Tokenizers Towards Fair Multilingual Language Modeling

Large language models rely on a single tokenizer that is chosen at training time and fixed thereafter. Most tokenizers are trained on English-dominant corpora, making them under-serve morphologically rich and non-Latin-script languages, where they can produce fragmented representations, longer sequence lengths, and reduced information density during training. We propose Mixture of Tokenizers MoTo, a modular tokenization framework that trains dedicated per-language BPE tokenizers and composes them into a unified superset vocabulary, where tokens shared across languages are deduplicated into a common ID space. Each tokenizer maintains its own normalization policy, pre-tokenization rules, and vocabulary budget. Despite allocating significantly fewer entries to English, MoTo retains English performance while delivering consistent gains across 18 out of 21 typologically diverse languages spanning six scripts, with improvements most pronounced on morphologically complex and non-Latin-script languages that are systematically underserved by English-dominant tokenizer design. Our results suggest that modular subword tokenization is a practical and extensible alternative to monolithic tokenization for multilingual parity.
Read the full abstract

Saturday 7 November

Number Theory

Character-Twisted Sums of Real Quadratic Class Numbers and a Selberg-Type Conjecture

Extending the arithmetic Selberg-zeta function framework studied by Sarnak \cite{Sarnak1982} and Hashimoto \cite{Hashimoto2007}, we study Dirichlet-character weighted sums of narrow class numbers of real quadratic orders and derive explicit formulas according to the scalar, split, ramified, and inert reductions of primitive hyperbolic classes. Our first main result establishes asymptotic formulas for the class number sums associated with the three noninert reduction types and shows that, after multiplication by the corresponding nontrivial even-character weights, their leading terms cancel exactly for every fixed prime $p\ge 5$. Computational experiments across the sampled ranges of levels and geodesic lengths support a substantially stronger bound for the sum over nontrivial even-characters and exhibit systematic cancellation among the scalar, split, and ramified contributions. If this conjectural bound persists over arbitrarily long polynomial ranges, the resulting exceptional eigenvalues-free regions approach the Selberg threshold $1/4$, providing a possible arithmetic route toward a Selberg-type spectral gap.
Read the full abstract

Metric Geometry

A Sharp Perimeter Inequality for Mean Distances on Convex Curves

For a planar convex body $K$, let $\theta(K)$ denote the expected Euclidean distance between two independent points uniformly distributed on $\partial K$ with respect to arclength. We prove the sharp inequality $$\frac{\theta(K)}{\operatorname{per}K}>\frac{1}{6}.$$ The constant is approached by rectangles degenerating to a segment.
Read the full abstract

Artificial Intelligence

How to make a moral agent: sentimentalism and rationalism as lenses on AI

Can AI systems be moral agents? That is, can they undertake actions that can be correctly assessed as morally right or wrong? There is no settled normative criterion of moral agency that applies neatly outside paradigm human cases. Because of the difficulties with a normative account, we instead propose a research program to explore descriptive moral agency. Most basically, we ask, what are the capabilities we humans possess that underlie our moral agency? To do so, we draw on historical scholarship in philosophy and moral psychology to isolate candidate capabilities. Then we suggest refining candidate capabilities so that they are testable in the context of AI systems. Finally, we offer specific empirical recommendations to test AI systems, especially large language models (LLMs), for those refined capabilities. Our approach offers several benefits. Firstly, it may demonstrate whether the behaviors of AI systems are aligned, or in a broad sense desirable. Secondly, the debate may inform policy questions (e.g., what models should be deployed in which contexts?). Thirdly, our approach may complement engineering and design efforts: when capabilities can be precisely characterized, they can inform the development of models. Fourthly, applying our moral reasoning concepts in the new context of AI systems should improve our understanding of those concepts.
Read the full abstract

Computer Vision

Technical Challenges for Ultra Fine Pitch Hybrid Bonding

Hybrid bonding is rapidly emerging as a foundational technology for next-generation advanced packaging, enabling ultra-high-density interconnects required for heterogeneous integration of chiplets, logic, memory, and specialized accelerators. As the industry pushes toward ultra-fine pitch interconnects approaching single- digit microns and below, significant technical challenges must be addressed across the process flow. These challenges include achieving extremely tight wafer and die alignment tolerances, maintaining surface planarity and cleanliness at the nanometer scale, controlling copper dishing and dielectric erosion during CMP, and ensuring defect-free oxide and copper-copper bonding interfaces. In addition, scaling hybrid bonding to ultra-fine pitch requires improvements in metrology, defect inspection, thermal management, and yield control to support high-volume manufacturing. Integration with complementary technologies such as silicon interposers, wafer-to-wafer and die-to-wafer bonding architectures, and heterogeneous material stacks further increases process complexity. This presentation reviews the key technical barriers to ultra-fine pitch hybrid bonding and discusses emerging process solutions and equipment innovations required to enable reliable, high-density interconnects for future high-performance computing, AI, and advanced system-in- package applications.
Read the full abstract

All authors Charles G. Woychik, Jonathan Abdilla, Chris Scanlan, Victor Vilar, Robert Patti, Dadi Setiadi, Himanandhan Reddy Kottur, Navid Asadizanjani

General Relativity and Quantum Cosmology

Resonance Induced Chaos of Charged Particles Around a Magnetized, Periodically Time-Conformal Reissner-Nordstrom Black Hole with Quintessence

Astrophysical black holes are never strictly static: accretion, mergers, and gravitational radiation drive a slow evolution of their geometry. We model this evolution through a periodic time-conformal deformation, $\Omega(t)=1+\epsilon\sin(\omega_d t)$, of a Reissner–Nordström black hole surrounded by quintessence, and study the resulting onset of chaos for charged test particles. The deformation converts the integrable static spacetime into a periodically driven nonlinear Hamiltonian system, so that the transition from regular to chaotic motion can be related to the breakdown of an approximate Noether symmetry. The resulting two-degree-of-freedom Hamiltonian system is solved using an adaptive high-order Runge–Kutta integrator, while the dynamical behaviour is analysed through Poincaré sections, bifurcation diagrams, the largest Lyapunov exponent, and the Smaller Alignment Index (SALI). Our results show that the onset of chaos is controlled primarily by nonlinear resonance rather than by the magnitude of any individual parameter. Chaos grows with the conformal amplitude $\epsilon$ through resonance overlap, while the magnetic field $B$ acts non-monotonically: it produces chaos near the field strength at which the radial frequency matches the drive, and suppresses it at stronger fields by detuning. The small-divisor structure of the approximate Noether symmetry predicts the magnetic threshold without free parameters ($B_c\simeq0.011$ against the measured $0.012$), and reproduces the conformal threshold once a single coupling constant is fixed. In contrast, the quintessence parameter mainly facilitates particle escape by reducing the confining potential barrier, providing a distinct non-resonant route to the breakdown of regular motion.
Read the full abstract

General Relativity and Quantum Cosmology

Finite Energy, Event Horizons, and a Universal Bound on the Singularity Order in Born–Infeld-type Black Holes

We investigate an interpolated Maxwell–Born–Infeld theory bridging the Maxwell and Born–Infeld limits. In the electrostatic sector the total energy is finite, and remains bounded above even as the theory approaches the Maxwell limit. We construct the electrically charged black hole, whose black hole solution interpolates monotonically between the classical Born–Infeld and Reissner–Nordström solutions; singularity can be relegated, independently of the weight parameters. For general interpolated theories with a truncation threshold, we analyze the existence and multiplicity of event horizons and classify the horizon configurations, and for general Born–Infeld-type nonlinear structures we prove a universal lower bound on the order of the Kretschmann divergence, thereby excluding regular electric black holes.
Read the full abstract

Computer Vision

Whole-Slice Misses Escape Round-Trip Filtering and Volumetric Dice in a Source-Free Flow-Matching Segmenter

SymmAdapt, a source-free flow-matching segmentation pipeline, relies on two quality gates: a round-trip filter on pseudo-labels and volumetric Dice in evaluation. We audit both on 120 AMOS CT patients unseen in training and adaptation. The model often misses the spleen on whole slices, severely so in 30.0% of patients. The round-trip filter passes 83.31% of these lost slices, because its round trip reproduces the miss and the filter scores only organs present in the pseudo-label. Volumetric Dice depends on how much volume is missed, not where, and self-training leaves 96.5% of lost slices lost (96.6% on a second noise stream). A model-free repair, Chain-and-Fill, met all pre-registered endpoints on 119 further patients and could not be distinguished from SAM 2 propagation in severe-group gain, with 1/357 rather than 63/357 harmed patient-seed pairs. Reporting lost slices next to volumetric Dice makes the error visible at no cost.
Read the full abstract