Searcharxiv⌕ Search

arXiv subjects

Abhishek Kumar

Publications and source records attributed to Abhishek Kumar.

At least 19 recordsLinked to original sources
arXiv2026

Baseline and Incremental: User, Item, and Request Debiasing in Instagram Feed Ranking

A feed ranker scores candidates with a value model, a weighted sum of engagement predictions. Each prediction conflates a baseline, what this viewer, item or request would produce anyway, with the incremental amount by which this candidate exceeds it. The literature mostly removes the baseline. We argue that a multi-signal value model should keep the raw prediction and add the incremental part as a separately weighted term. On Instagram Feed, three debiasing families deployed independently, for viewer habit, item popularity and request context, are one parameterised operator, and a design procedure picks a point in it. Media-type (format) bias goes instead to a debiased label chosen by the same procedure. In a controlled four-arm ablation of deployed Feed terms (neither term, each alone, both), the two-term arm beats both single-term arms at production weights on point estimates in 11 of 13 judged (term, metric) pairs. Of the 26 direct contrasts, 17 are significantly for it and only 1 against it, on impressions. The ablation also suggests that the offline diagnostic should screen signals, not rank them. Such terms have served Instagram Feed production for most of a year. Beyond Feed, Meta, whose apps serve billions of people, has adopted the operator across surfaces, apps and both ranking stages. Its launches there drive sessions and daily active users, the core visitation and retention metrics.

cs.IR↗
arXiv2026

Floquet-ADAPT-VQE: A Quantum Algorithm to Simulate Non-Equilibrium Physics in Periodically Driven Systems

Periodically driven quantum systems exhibit many fascinating phenomena absent in equilibrium systems, but their simulation is more challenging than that of static systems. Consequently, quantum simulation of these systems offers greater opportunity for achieving quantum advantage. To build the foundation for simulating time-periodic Hamiltonians, we utilize the Floquet-Hilbert space formalism to transform the Hamiltonian into a time-independent form and provide its Pauli decomposition. Within this formalism, we develop a hybrid quantum-classical algorithm, Floquet-ADAPT-VQE, that produces three key quantities that govern periodically driven quantum dynamics: the quasienergy, the associated Floquet eigenstates, and expectation values of observables with respect to these eigenstates. The latter is achieved by applying a fixed-depth quantum circuit to an initial Floquet eigenstate---an efficient construction made possible by the Floquet-Hilbert space formalism. We also identify a suitable auxiliary initial state whose squared Floquet energy is independent of the number of auxiliary qubits as well as the driving frequency, which leads to better convergence with fewer ADAPT iterations. We demonstrate our algorithm by performing numerical simulations on a periodically driven XYZ model with a magnetic field. We also explore potential applications of our algorithm for studying various non-equilibrium phenomena in periodically driven systems.

quant-ph↗
arXiv2026

From Self-Distillation to Self-Practice: Privileged Information for Multi-Turn Agents

On-policy self-distillation (OPSD) has become a popular recipe for post-training LLM agents. It supervises the agent model at the token level with a stronger teacher view of the same model, obtained by conditioning on privileged information (PI). In this work, we show that in multi-turn agents, this paradigm teaches the student to act with confidence but without the information behind it. The trained agent behaves as if it had privileged information it never observed, and its performance falls well short of plain RL, in the worst case below the untrained base model. Therefore, we propose Privileged Self-Practice (PSP), which keeps the PI and moves it from the loss to the sampler. When the student's rollouts on a task mostly fail, we inject a short per-task instruction written by an analyzer model, sample the task again with the instruction in context, and train on the result with an unchanged GRPO objective. The privileged information stays in the prompt and never enters the loss. Across AppWorld and SWE-bench Verified, with three different student models, PSP obtains the best average score in every setting and is the only method that consistently outperforms plain GRPO, with relative improvements over GRPO of 7% to 58% in average score across settings. Project page: https://sux1ngyu.github.io/PSP/.

cs.AI↗
arXiv2026

CoMIC: Collaborative Memory and Insights Circulation for Long-Horizon LLM Agents in Web Cloud-Edge Systems

LLM-based Web agents are increasingly deployed on edge servers to carry out long-horizon tasks on behalf of users with low latency. However, the lightweight models that fit on edge servers often struggle with long-horizon tasks that require persistent memory, subgoal tracking, and reflection. Fine-tuning edge models after deployment is costly and difficult to scale across heterogeneous nodes, while purely local execution confines each agent to its own isolated experience. We propose \textsc{CoMIC}, a parameter-update-free cloud-edge framework for Collaborative Memory and Insights Circulation. Following the paradigm of ``Act at the Edge, Reflect in the Cloud'', edge agents execute locally with subgoal-oriented hierarchical memory and selective re-expansion of relevant histories, while a cloud-side LLM critic asynchronously evaluates the trajectories uploaded through a Web gateway, admits reusable experience, and circulates cross-agent guidance indexed by semantic subgoal identifiers back to all edge agents. On five long-horizon benchmarks, \textsc{CoMIC} enables weak edge agents to achieve non-zero success on tasks where a standard memory-based agent fails entirely, and improves their task progress over both the standard agent and purely local execution while reducing context consumption relative to the standard agent, all without updating model parameters. When stronger heterogeneous agents join the edge, the weak agents further improve in success, progress, and grounding with lower context consumption.

cs.AI↗
arXiv2026

An LLM-powered Agentic Recommendation System for Connected TV Content Discovery

Recommendation systems, from traditional multi-stage to recent unified generative architectures, face challenges in incorporating diverse contextual signals, such as trending topics, breaking news, cultural events, and cross-surface user activities, into their ranking pipelines. These systems are designed to consume structured behavioral signals with consistent schemas, and lack the reasoning capability to naturally process unstructured or heterogeneously formatted contextual information. Incorporating such signals typically requires feature engineering, bespoke data pipelines, and carefully tuned heuristics. In this paper, we present an LLM-powered agentic recommendation system designed for Connected TV (CTV) content discovery that addresses these limitations. Our system leverages the reasoning capabilities of large language models to naturally process and synthesize diverse signals across varying schemas and structures, eliminating much of the manual integration inherent in traditional ranking and retrieval systems. Recognizing that current LLM-based solutions still fall short of traditional machine learning models in several recommendation tasks, including retrieval efficiency, personalization precision, and scalability, we adopt an agentic architecture that orchestrates specialized components, allowing each sub-task to be handled by the most suitable method, whether LLM-based or traditional ML. The main contribution of this work is our engineering approach to successfully overcoming the practical limitations of enabling LLM for recommendation, particularly inference latency. We share insights from our work and discuss the trade-offs and lessons learned in building a hybrid system that combines the flexibility of LLMs with the performance of established recommendation techniques.

cs.IR↗
arXiv2026

Know What You Don't Know: Agentic Data Synthesis by Evolving Generation Space for Closed-Loop Learning

The continued improvement of large language models increasingly depends on synthetic experience, yet effective data synthesis for interactive agents requires not only executable and verifiable trajectories, but also interactions that provide learning value for the current task agent. Existing methods increasingly incorporate environment feedback, data validation, or learner failures, but still face a feedback-to-decision gap: observed weaknesses do not directly specify how future synthesis should change, or whether they can be represented by the current synthesis space. To address this gap, we propose EvoSyn, a learning-feedback-driven multi-agent framework that organizes synthesis around a revisable state specifying what interaction factors can be deliberately generated and where synthesis effort should be allocated. A role-specialized multi-agent system translates learner feedback into synthesis revisions, while a Designer allocates synthesis directions and a Planner grounds them into executable experience. This process enables within-space adaptation toward weak or under-covered regions and synthesis-space expansion that promotes newly discovered factors into explicit generation controls. Experiments on $τ^2$-Bench and CAR-Bench show that EvoSyn improves average task success by 13.3 points over the strongest synthesis baseline, while producing higher-quality and structurally broader synthetic experience and discovering controllable factors missed without learner feedback.

cs.AI↗
arXiv2026

MLCommons Jailbreak Benchmark v1.0

Modern AI systems are designed to refuse hazardous requests. A jailbreak is a prompt crafted to bypass those safeguards and elicit outputs that the system would normally refuse to provide. The MLCommons Jailbreak Benchmark v1.0 provides an end-to-end methodology for evaluating the robustness of large language models to single-turn, text-based jailbreak attacks. It combines criteria-driven system and attack selection, paired baseline and adversarial evaluation, human annotation, automated evaluator calibration, scoring, grading, and risk-calibrated disclosure within a single benchmarking pipeline. The benchmark evaluates eight open-weight systems using 264 seed prompts spanning eleven hazard categories and representative attacks drawn from the MLCommons Jailbreak Taxonomy. Responses are assessed using the AILuminate Assessment Standard v1.4, and robustness is measured through the Resilience Gap: the change in safety performance between baseline and adversarial conditions. Across all evaluated systems and attacks, the unsafe-response rate increased from 11.08% under baseline conditions to 18.65% under jailbreak conditions, producing an average Resilience Gap of 7.57%. Accessible systems showed a larger mean gap, while attack effectiveness varied substantially across attack categories and hazards. The benchmark also examines evaluator reliability and sources of measurement error. Beyond reporting results, Jailbreak Benchmark v1.0 establishes a reproducible methodological foundation for comparative jailbreak evaluation and for future expansion across systems, attacks, hazards, and evaluation methods.

cs.AI↗
arXiv2026

Inverse Design of the Topology Bandwidth Tradeoff in Valley Photonic Crystals

Valley photonic crystals provide a route to broadband domain-wall guiding, but increasing the bulk gap can delocalize Berry curvature and reduce the non-quantized valley-Chern number. We introduce a mixed-integer particle-swarm framework that searches polygonal VPC unit cells and maps the Pareto tradeoff between normalized bandgap and valley topology using the bulk objective $\mathcal{T}=(Δf/f_0)^2|C_v|$. The selected designs are then evaluated independently using full-wave transport simulations through disorder-averaged propagation-length extraction in $200a$ domain-wall waveguides with positional disorder $σ=0.01a$. Across the studied design ensemble, the extracted propagation length exhibits a positive empirical association with $|C_v|$, providing a direct multi-design connection between the patch-integrated valley-Chern number and quantified disorder-limited transport in this VPC family. The framework therefore treats bandwidth, valley topology, and device robustness as distinct but connected quantities, enabling balanced inverse design rather than optimization of a single transport metric alone.

physics.optics↗
arXiv2026

GA-EIRFS: A Geometry-Augmented Repeat-Factor Sampling Method for Long-Tailed LiDAR 3D Object Detection

Long-tailed 3D object detection is treated as a class-frequency problem, but LiDAR supervision quality depends on object observability: similar frequencies can hide different geometric evidence. We introduce Geometry-Augmented Exponentially Weighted Instance-Aware Repeat Factor Sampling (GA-EIRFS), a detector-agnostic method that modulates a frequency-based repeat factor with a fixed geometry score combining point count, surface-normal entropy, and surface coverage. GA-EIRFS changes only frame-sampling probabilities, leaving the detector and inference unchanged. On nuScenes it improves mean average precision (mAP) and the nuScenes detection score (NDS) in four converged experiments with CenterPoint and PointPillars over two seeds; for CenterPoint at seed 666, mAP rises from 0.552 to 0.563 and bicycle AP from 0.306 to 0.359. Per-class gains correlate with the class sampling-weight increase (Spearman rho=0.70, p=0.025) but not with geometry score alone (rho=0.32, p=0.37), so geometry amplifies frequency-driven need. KITTI results vary across seeds, most for the rarest class. Code: https://github.com/Multimodal-Sensing-Lab/GA-EIRFS.

cs.CV↗
arXiv2026

Persistent Negatives for Adversarial Black-Box On-Policy Distillation

Black-box On-Policy Distillation (OPD) seeks to improve a student from its own generations when the teacher provides sampled responses but not token probabilities. Adversarial distillation offers one route: it learns a discriminator over prompt-matched teacher and student responses and uses its score as the policy reward. However, sampling discriminator negatives from the latest student at each step couples the learned reward to a negative distribution that changes after every policy update. We address this moving-target problem with persistent-negative adversarial distillation, a live-pool method that replaces a fraction of each discriminator batch with historical, prompt-matched teacher--student comparisons. Under matched discriminator compute, historical comparisons train the discriminator, while GRPO remains on-policy with fresh student responses. Our analysis identifies the Bayes-optimal reward as a teacher-to-negative log-density ratio and, under explicit assumptions, shows how persistent negatives anchor the discriminator and reduce reward-estimation MSE relative to fresh-negative training. Across two student families, three judges, and four judged-chat benchmarks, persistent-negative adversarial distillation consistently improves performance over current methods at matched discriminator compute. It also yields smoother fresh-policy discriminator trajectories, with fewer below-chance dips. These findings identify the discriminator's negative distribution as an important design axis in black-box on-policy distillation.

cs.CL↗
arXiv2026

CP-Agent: A Harness-Engineered Agent for Crystal Plasticity Simulation Workflows

Crystal plasticity (CP) simulations predict the mechanical behavior of polycrystalline metals, yet their routine use is hindered by the manual effort of configuring heterogeneous tools, orchestrating multi-step data pipelines, and calibrating constitutive parameters against experiments. These bottlenecks impede productivity in systematic parameter studies, motivating interest in automated workflows. This study presents CP-Agent, a harness-engineered LLM-based agent that autonomously executes complete CP modeling workflows from natural-language tasks. Operating under the ReAct paradigm, the agent reasons about tool selection and sequencing while delegating numerical search to established optimizers. The harness comprises a minimal system prompt, typed tool definitions, a dispatcher, and a safety-bounded iteration loop, encoding domain knowledge through tool schemas rather than hard-coded logic. CP-Agent is demonstrated on four case studies: calibrating four slip parameters of additively manufactured stainless steel 316L against tensile data; validating the workflow against published copper benchmarks, reproducing stress-strain and texture evolution; recovering the initial crystallographic texture of copper, where the agent correctly identifies a diffuse initial texture; and reproducing the multi-pass rolling texture evolution of a Mg-Zn-Ca alloy, where the agent chains five deformation passes and recovers the experimentally observed weakened, split basal texture. In all cases, the agent inferred the correct execution sequence from the task statement, robustly across repeated runs, and delivered physically interpretable results. This work establishes harness engineering as a systematic approach to automating CP modeling workflows while maintaining physical interpretability and auditability through visible reasoning traces.

cs.AI↗
arXiv2026

A 39pJ/b 7.3Gbps 1.3mm$^2$ Multi-Subcarrier Massive MU-MIMO-OFDM Detector Exploiting Beamspace Sparsity and Frequency-Domain Correlation in 22FDX

We present the first multi-subcarrier massive multi-user (MU) multiple-input multiple-output (MIMO) orthogonal frequency-division multiplexing (OFDM) data detector reported in the open literature. By exploiting the channel's beamspace sparsity and frequency-domain (FD) correlation, we achieve up to 3x area and power reduction. Our design supports U=8 user equipments and B=64 basestation antennas, computes soft outputs for QPSK to 256-QAM, and processes 16 subcarriers in parallel. The fabricated 22FDX ASIC has a core cell area of 1.3mm$^2$, consumes 286mW, and delivers a throughput of 7.3Gbps at 0.8V core voltage, achieving best-in-class energy efficiency of 39pJ/b.

eess.SP↗
arXiv2026

SE-MoLoRA: Shared-Expert LoRA Adapters for Domain-Specific Photographic Assessment

Vision-language models can describe images fluently, but they often fail to provide actionable photographic critique because semantic content and aesthetic judgment remain entangled. We propose SE-MoLoRA, a modular parameter-efficient adaptation framework for domain-specific photographic assessment. The method separates general photographic knowledge from specialist residual judgments using an always-active shared LoRA expert and routed adapters for composition, lighting, and technical quality. A lightweight query router selects the relevant specialist, enabling targeted critique without training separate full models. A rank-64 shared adapter captures broad photographic vocabulary, while rank-32 specialists learn domain-specific residuals with an orthogonal regularization penalty that encourages disentangled representations. Training data is obtained by distilling the Reddit Photo Critique Dataset into domain-labeled critique samples. On held-out critique generation, SE-MoLoRA improves BERTScore-F1 from 0.2317 to 0.4215 over monolithic LoRA and is preferred in 84.6\% of pairwise comparisons, while using fewer active parameters than separate specialist models. SVD-based ablation study shows that shared-specialist decomposition and orthogonal regularization reduce expert overlap. These results demonstrate that modular adaptation improves controllability and specificity in multimodal photographic critique.

cs.CV↗
arXiv2026

Field-Localized Forgery Detection for Digital Identity Documents

Digital onboarding and eKYC systems used by banks, fintech platforms, telecom providers, and other third-party services commonly verify users by comparing an uploaded identity document with a selfie or live facial capture. This workflow is convenient, but it also makes verification systems vulnerable to localised document manipulations, such as replacing the facial photograph, editing textual identity fields, or altering both. Existing image-forgery detectors are largely designed for natural images and do not explicitly account for the structured layout of identity documents, where security-relevant information is concentrated in specific semantic fields. We propose FLiD, a lightweight field-localised framework for identity-document forgery detection. Instead of processing the full document image, FLiD localises the facial and textual regions using a fine-tuned YOLO11 detector, extracts compact representations with a frozen MobileNetV3-Small backbone, and classifies field-level forgeries using a small 191K-parameter head. Face and text detectors are trained independently and combined through score-level fusion for documents containing simultaneous manipulations. Under 5-fold cross-validation, FLiD achieves AUC scores of 0.834, 0.926, and 0.837 for face, text, and combined attacks, respectively, while reducing Equal Error Rate by 28--29 percentage points compared with a full-document baseline. FLiD also outperforms general-purpose forgery detectors while requiring 13x fewer trainable parameters and 21x fewer FLOPs per field (~8x per full multi-field document), making it a promising candidate for data and resource constrained KYC deployment.

cs.CV↗
arXiv2026

Symmetry Breaking by Interfacial Dead Layers: Observation of Forbidden Self-Induced Spin-Orbit Torque in Symmetric Ferromagnets

Conventionally, spin-orbit torques (SOTs) in ferromagnets require heavy-metal layers or engineered structural asymmetry to break inversion symmetry. In this work, we report the observation of robust, self-generated SOTs in a nominally symmetric, heavy-metal-free MgO/NiFe/MgO trilayer - a geometry where such torques are theoretically forbidden. By combining harmonic Hall measurements with SQUID magnetometry and X-ray photoelectron spectroscopy, we identify the symmetry-breaking origin: a 1.8 nm magnetic dead layer at the bottom interface. Crucially, we demonstrate a quantitative agreement between our data and the drift-diffusion theory predicted by Kim and Lee, yielding a theoretically extracted dead-layer thickness (1.2 nm) which matches structural characterization. Furthermore, density-functional calculations confirm that NiFe possesses sufficient intrinsic spin Hall conductivity to support the observed spin currents. These results reframe the parasitic dead layer as a functional spintronic component, establishing a universal, all-ferromagnetic route to SOTs in standard magnetic heterostructures.

cond-mat.mes-hall↗
arXiv2026

Universal crossovers in weakly-monitored quantum critical states

We study post-measurement ensembles of ground states of tricritical and critical 1D quantum Ising Hamiltonians subjected, respectively, to weak energy and spin measurements without post-selection. These measurements act as relevant perturbations about the unmeasured critical ground states. Using finite-size renormalization group (RG) crossover analyses, we characterize their universal properties through the entanglement effective central charge, effective Affleck-Ludwig boundary entropy, and signatures of multifractality from moments of measurement-averaged correlation functions. In both cases, we find evidence for "measurement-dominated" or "measurement-altered" fixed points governed by the underlying Born-rule randomness. For critical Ising, we find a direct RG flow to a projective-measurement fixed point with area-law entanglement, whereas for the tricritical Ising model, we find evidence for a weak-measurement fixed point with logarithmic entanglement. These results clarify the RG-flow structure of weakly measured multicritical Ising ground states and show how intrinsic measurement-induced randomness can generate complex and rich universal long-distance scaling behavior in the post-measurement ensembles, accessible to controlled analytical RG and numerical finite-size RG crossover analyses.

cond-mat.stat-mech↗
arXiv2026

Harnessing Native Chromium Oxidation for Giant Orbital Torque and Field-Free Magnetization Switching in NiFe/Cr Bilayers

Orbital currents offer charge-to-spin conversion beyond the efficiency limit of conventional heavy-metal Spin Hall sources. However, harnessing them has so far required either thick orbital-Hall materials or additional heavy-metal conversion layers. Here, we show that the native oxide of chromium, typically regarded as parasitic, transforms a simple NiFe\Cr bilayer into a self-contained dual-channel orbital-current source without the need for any conversion layer. First-principles calculations predict a nearly threefold enhancement of the orbital Hall conductivity upon surface oxygenation, driven by Cr(3d)-O(2p) hybridization. Experimentally, naturally oxidized NiFe\Cr heterostructures exhibit a giant damping-like torque efficiency of $3.9 \times 10^{6}$ $Ω^{-1}$ m$^{-1}$, exceeding Pt (Ta) by one (two) orders of magnitude. The torque depicts a non-monotonic Cr-thickness dependence which cannot be explained by a conventional model. We have developed a drift-diffusion model with an oxidation-gated interfacial source which quantitatively reproduces the data, revealing that the Cr-CrO$_x$ interface generates orbital currents over an order of magnitude stronger than the bulk orbital Hall channel with an orbital transport length of $\approx 4$ nm. The enhanced torque enables field-free magnetization switching at $1.58 \times 10^{11}$ A m$^{-2}$, outperforming heavy-metal and CuO$_x$ benchmarks. These results establish native oxidation as a scalable strategy for realizing efficient orbital-torque devices.

cond-mat.mes-hall↗
arXiv2026

Element-Specific Visualization of Layer-Parity and Twist-Dependent Magnetism in CrSBr

Van der Waals (vdW) based antiferromagnets (AFMs) are an ideal platform for probing and understanding thickness- and twist-angle-dependent emergent spin phenomena. However, element-specific nanoscale characterization of the spin structure in atomically thin vdW-based AFMs systems and layer-parity effects remain elusive, making them crucial for both fundamental insight into low-dimensional magnetism and the rational design of spintronic devices based on these materials. Here, we utilize X-ray magnetic circular and linear dichroisms paired with photoemission electron microscopy to resolve the magnetic order in atomically thin CrSBr. Our comprehensive measurements reveal CrSBr magnetic structure at the nanoscale and its dependence on the layer number, surface encapsulation, temperature, and applied field. Moreover, in the orthogonally twisted bilayer configuration, obtained by twisting two CrSBr ferromagnetic monolayers by 90$^\circ$, the magnetic easy axis fundamentally differs from the individual monolayers, unlocking a new pathway for moiré magnetism.

cond-mat.mes-hall↗