Machine Learning · Cross-listed from cs.CL
Prediction Dynamics in Depth-Recurrent Language Models
Xinyue Luo and Fei Yu
Depth-recurrent
language models refine predictions through repeated latent updates. Why can intermediate answers agree with the endpoint while their scores continue to change? We derive a sharp margin characterization that decomposes the conservatism of a magnitude bound into common translation, direction relative to the winner, and the pairing of each competitor's update with its score gap. Across Huginn-3.5B and Ouro-1.4B, accounting for update direction and competitor pairing reduces the mean earliest qualifying depth by a further 22.5-34.4% of the total depth beyond translation removal under full answer-text scoring. This retrospective comparison uses completed trajectories. Substantial contributions also occur under label scoring. For shared predictive distributions, we separate common and contrast motion orthogonally and express the common component through candidate-set mass and within-set concentration. Common and contrast energies can attenuate at different rates, allowing a growing preference-change share to coexist with shrinking absolute updates. These findings explain finite-depth answer preservation through the geometry and composition of observed score changes.
Read the full abstract ▾
Machine Learning
Probabilistic Forecasting of Business Process Executions with Neural Temporal Point Processes
Jiaxin Yuan, Daniela Grigori and Han van der Aa
Operators of service-based systems act on forecasts of how a running execution will continue, and such a forecast is actionable only if its reliability is known. Mainstream deep-learning models for this task are discriminative and deterministic: they emit a single next activity and a single remaining-time estimate, without a distribution to reason over. We instead cast the problem as generative sequence modelling with marked temporal point processes, which define a joint density over the next mark and its inter-event time and therefore deliver predictive distributions by construction. Real event logs violate the simple-point-process assumption these models rest on, since consecutive events frequently carry identical timestamps; we handle such ties explicitly and combine a
transformer encoder with a mixture decoder over inter-event times, trained by exact log-likelihood. On ten public logs, the resulting model matches discriminative baselines on point accuracy, dominates them on the calibration and sharpness of remaining-time distributions, and is the cheapest at inference, since a full predictive distribution is obtained in a single forward pass without sampling.
Read the full abstract ▾
Machine Learning
Knowledge-Graph-Augmented Chronos-2 for HEC-RAS Surrogate Forecasting
Edward Holmberg, Elias Ioup and Mahdi Abdelguerfi
We investigate whether coupling a time-series foundation model to hydraulic project knowledge improves surrogate
forecasting of HEC-RAS water-surface elevation (WSE). We present KG-Chronos-2, which combines a frozen Chronos-2 predictor with exact-state residual decoding, graph-conditioned historical
retrieval, and input-aligned correction. We compare the method with persistence, a residual LSTM, project-conditioned recurrent GeoFNO, a hydraulic DCRNN-style model, and frozen Chronos-2. Task-specific fitting uses the 2008 simulation. Evaluation covers 64 fixed 24-hour windows from the 2011 and 2002 simulations at 4,675 cross sections in 71 reaches on a shared geometry. KG-Chronos-2 achieves event-balanced root-mean-square error 0.246970 in native WSE units. It reduces RMSE by 14.13% relative to frozen Chronos-2, 29.38% relative to the hydraulic DCRNN-style model, and 39.54% relative to recurrent GeoFNO. The 95% hierarchical-bootstrap interval for its event-balanced RMSE difference from frozen Chronos-2 is [-0.075177, -0.016317]. KG-Chronos-2 also achieves the lowest active-window and final-lead RMSE among the six completed systems. These results support coupling a frozen temporal predictor to project knowledge for warm-start HEC-RAS
forecasting on the fixed
benchmark.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.CV
Hiding in Plain Sight: A Diffusion-based Mitigation of Geolocation Privacy Leakage in Vision-Language Models
Yining Wang, Xi Li, Mi Zhang et al. (+4 more)
Multimodal large
reasoning models (MLRMs) have demonstrated remarkable capabilities in complex visual understanding. However, this very power introduces a critical yet underexplored
privacy threat: adversaries can exploit MLRMs to precisely infer users' geographic locations from casually shared photographs, by performing structured
reasoning over subtle visual cues such as architectural styles, vegetation, and lighting conditions. In this work, we present a systematic study of MLRM-driven geolocation
privacy leakage. We first reveal that refusal-based safeguards are critically insufficient, as carefully crafted jailbreak
prompts can raise model response rates to 100%. We further identify that existing defenses, which inject imperceptible perturbations into shared images, suffer from structural limitations intrinsic to their pixel-space optimization, resulting in degraded black-box transferability and pronounced visual artifacts. Motivated by these findings, we propose a
diffusion-based framework that provides targeted, proactive defense against geolocation
privacy leakage. By injecting perturbations into the latent space of a
diffusion model during reverse sampling, our method operates directly on high-level semantic representations, thereby resolving the effectiveness-utility bottlenecks by construction. We further ground our optimization with GeoCLIP, a model explicitly aligned with GPS coordinates, as a surrogate to pinpoint and disrupt the geographic signals that MLRMs exploit for location inference. This targeted semantic disruption yields significantly stronger black-box transferability while preserving perceptual image quality, offering a seamless integration on social media platforms. Code is available at https://github.com/RachelWolowitz/Hiding_in_plain_sight.
Read the full abstract ▾
All authors Yining Wang, Xi Li, Mi Zhang, Xiaohan Zhang, Xiaoyu You, Zhenxing Qian, Mi Wen
Machine Learning
IntBMoE: Integrating Block-Level Conditioning into Expert Composition for Full-Participation Mixture-of-Experts
Ran Cheng, Longfei Xu, Zheng Liu et al. (+2 more)
Mixture-of-Experts (MoE) scales capacity, but existing designs cannot set three quantities independently. For a single
token, participation is how many experts contribute knowledge to its output, execution is how many are actually computed (compute cost), and materialization is how many expert-sized parameter sets must be built and stored (
memory cost). Sparse routing keeps execution and materialization low, but shrinks participation: for each
token, only a few experts contribute. Dense output-mixing restores full participation, but its execution grows with the number of experts. Parameter-merging keeps execution at one expert, but its materialization grows with the number of routing decisions. We propose IntBMoE, a block-conditioned MoE that decouples all three by pairing dense expert composition with sparse block execution. Its blocks come from a small learned codebook, one per entry. At each internal layer, a lightweight hypernetwork merges all expert bases in that layer's pool into one composed expert. Participation is full, because every composed expert draws on the entire pool. Execution stays sparse, because a router sends each
token to only a few blocks. Materialization is bounded, because the codebook, not the input, fixes how many blocks exist. Dual-Path Residual Gating (DPRG) further couples two independently composed paths through multiplicative gating. Experiments on image classification show consistent gains over representative sparse and dense MoE baselines. Additional experiments on
language modeling and sequential recommendation validate its generalization beyond
vision. IntBMoE is fully deployed in AMap's generative recommendation system, serving hundreds of millions of users under a 60ms latency budget, with a 2.4% relative UVCTR gain in online A/B testing. Our code is available at https://github.com/AMAP-ML/DreamX-Rec/.
Read the full abstract ▾
All authors Ran Cheng, Longfei Xu, Zheng Liu, Kaikui Liu, Xiangxiang Chu
Machine Learning
Routine Blood Tests Outperform CRP for Distinguishing Bacterial From Viral Infection in Children
Mihaela Demireva, Zhecho Mitev, Djuna Chinareva-Klimentova et al. (+3 more)
Acute infectious diseases are among the leading causes of medical consultations and hospitalizations in children worldwide. These infections are predominantly caused by viruses or bacteria, yet differentiating between the two remains a common clinical challenge. As a result, pediatricians often default to the safer option of prescribing antibiotics contributing to the growing problem of antimicrobial resistance. The objective is to assess the additional predictive value of CBC towards determining the current infection. This retrospective study used data from 906 pediatric patients aged between 2 and 14 years who were tested positive either for viral or bacterial infection between 2022 and 2026. Inclusion criteria further required availability of CBC results and CRP level measurements. These laboratory parameters as well as age were used as input features for several supervised classification models. Model performance was evaluated using AUC, sensitivity and specificity. The best performing model is XGBoost, which included all features, achieving out of-sample performance of AUC of 81.7% and sensitivity of 70.8%, specificity of 79.2%. All trained models outperform a CRP-based only decision-rule model in terms of AUC. We suggest that the decision to prescribe antibiotics should be based on a number of factors, including but not limited to CBC, some of which are not currently incorporated into routine practice.
Read the full abstract ▾
All authors Mihaela Demireva, Zhecho Mitev, Djuna Chinareva-Klimentova, Svetoslav Ivanov, Georgi Nalbantov, Dimitar Mitev
Machine Learning
Deep Reinforcement Learning with Buffered Quantile Objectives
Mohammad Alipour-vaezi and Sajad Khodadadian
Quantile-based reinforcement learning provides an interpretable approach to risk-sensitive decision-making by optimizing a prescribed
quantile of the cumulative-return distribution. Despite this appeal, learning under a point
quantile objective is challenging:
quantiles can change abruptly under small perturbations of the return distribution, and exact
quantile-sensitive planning requires computationally demanding distributional optimization. Lower-buffered
quantiles alleviate the former difficulty by averaging neighboring
quantiles immediately below the target level, providing a smoother surrogate while preserving the underlying point-
quantile objective. Existing methods based on this principle, however, remain model-based and rely on explicit return-law planning, limiting their applicability beyond small tabular problems. We develop Deep-BQRL, a model-free distributional reinforcement-learning framework that extends buffered-
quantile learning to neural function approximation. The method learns conditional return
quantiles directly from sampled transitions, constructs buffered action scores from the relevant region of the learned
quantile function, and uses ensemble disagreement to guide exploration. An augmented input representation allows the learned policy to respond to trajectory information without explicitly reproducing the
quantile-state recursion required by exact planning. Experiments on an asset-selling optimal-stopping problem and slippery FrozenLake compare Deep-BQRL with model-based UCB-BQRL and tabular PPO and TRPO implementations. In asset selling, Deep-BQRL attains smaller mean cumulative point-
quantile policy gaps than PPO and TRPO at the reported target levels, while UCB-BQRL retains the smallest gaps. The learned stopping decisions also vary with the target
quantile, providing an interpretable illustration of the method's risk-sensitive behavior.
Read the full abstract ▾
Machine Learning · Cross-listed from stat.ML
Sparse Identification for Automatic Large-Scale Screening: A Constraint-Aware Framework with Ultra Fast Decoding Algorithm
Jianing Li, Li Chai and Yingcheng Lai
In the early stages of a pandemic, identification of a small number of infected individuals through large-scale screening is critical for pandemic control, yet remains challenging under limited reagents and testing capacity. Existing group testing methods suffer from either high computational complexity or low identification accuracy. Even worse, no available methods provide theoretically rigorous analysis for sparse identification with hard constraints caused by the sample usage constraint and the dilution effect existing ubiquitously in practical applications. In this article, we propose the Logic Screening method (LoSc), an ultra fast, accurate, and theoretically grounded framework for large-scale screening. LoSc introduces a novel decoding
algorithm with a very simple selection strategy, achieving identification of all positives with only O(klogn) pooled tests. The decoding relies only on logical operations, enabling direct hardware implementation and yielding ultra fast computational implementation. Moreover, LoSc explicitly incorporates dilution and sample usage constraints into pooling designs, and establishes theoretical guarantees to guide optimal pooling configurations. Extensive simulations confirm the superior effectiveness, efficiency, and scalability. We believe LoSc offers a fast and reliable solution for automatic large-scale screening.
Read the full abstract ▾
Machine Learning · Cross-listed from stat.ML
Diagonalized Attention for Individualized Regression: Latent-Row Localization and Prediction
Borui Peng, Liwei Lin, Feifei Wang et al. (+1 more)
Modern text and image representations are often
matrix-valued, with rows corresponding to
tokens, patches, or other local feature vectors. Predictive information is often sparse but sample-specific, making classical sparse regression methods with a common support poorly suited to this heterogeneity. This paper formalizes an individualized sparse regression framework for
matrix-valued
covariates in which each observation has its own rows of interest, while the associated regression effects are shared across the population. To estimate this model, we introduce a diagonalized
attention mechanism that uses query–key scores to localize sample-specific signal rows and a value
matrix for downstream regression. The proposed method has a parameter dimension independent of sample size and can identify rows of interest for new observations without their responses. We establish existence
theorems showing that, under suitable score-separation and concentration conditions, single-head and multi-head diagonalized
attention models recover the latent rows with high probability, yielding prediction risk bounds. Our theory therefore provides a statistical explanation of how
attention-based scoring localizes sample-specific signals in heterogeneous
matrix-valued data. Simulations demonstrate strong prediction and localization in regression and misspecified classification across varying sample sizes, dimensions, and signal cardinalities. Real sentiment analyses show improved classification accuracy and interpretable
token selection.
Read the full abstract ▾
All authors Borui Peng, Liwei Lin, Feifei Wang, Long Feng
Machine Learning
An Introduction to Compression-Based Machine Learning
John Hurwitz, Edward Raff and Charles K. Nicholas
Any lossless compression
algorithm (like gzip) may be converted into a machine learning method, via either Normalized Compression Distance or the Minimum Description Length principle. Any auto-regressive model may be converted into a lossless compression method via
entropy coding. This seemingly circular dependence has unrealized potential in modern artificial intelligence and machine learning, and we survey and formalize the various strategies that have been used to leverage compression for machine learning. We introduce and empirically validate a design framework for compression-based ML, finding compression-based methods competitive with conventional baselines and decisively stronger on malware. We find that varying these design choices yields accuracy gains of up to 0.62.
Read the full abstract ▾
Machine Learning
Fast And Accurate Text Content File Type Identification
Manu Nandan, Michael Brautbar and Edward Raff
A common requirement across organizations is to have a tool that can identify file types based on their contents, particularly in the cybersecurity domain where magic numbers and file extensions can not be trusted. While existing tools work well in practice, there is plenty of room for improvement either in terms of computational load and time for detection in the case of model based tools like Magika or in terms of accuracy of detection in the case of file parsing tools that use programming
language constructs. In this study, we propose a neural network model for identification of types of text content files, especially source code, that is more accurate and faster than other available tools. Our experiments on open-source files indicate that it is not only more accurate on average for text-content file-type identification, but also approximately four times faster than Magika, while being 28% smaller in size.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.CR
Identifying Security Platform Product Abuse with Machine Learning
Shaefer Drew, Michael Brautbar, Paul Knight et al. (+6 more)
Product abuse is an individually rare, but growing, problem across the SaaS industry. Highly sophisticated threat actors can misuse security platforms within customer environments or conduct bypass experiments on the product itself. Threat actors can leverage living-off-the-land (LOTL) attacks to avoid using cumbersome, frequently detected malware. Remediating this threat requires collecting multiple data modalities across different types of databases, addressing a cold-start problem in the intrinsic rarity of such sophisticated but dangerous events, and designing within the constraints of real-world deployment (e.g., cost, user behavior, performance, etc). To wit, we provide the first study of such a whole-system defense, especially with respect to a deployed and operational capability. Our results show an increase in product abuse coverage by 35%, a 30% reduction in monthly alerts, and adaptability to changes in malicious actors' behavior. We review both the constraints we considered in designing the system to meet operational requirements and a retrospective evaluation of the value of explainable features and counterfactual performance on previously identified attacks.
Read the full abstract ▾
All authors Shaefer Drew, Michael Brautbar, Paul Knight, Edward Raff, Lana Peric-McDermott, Simran Sarin, Nickolas Machado, Hanna Albright, Vitaly Zaytsev
Machine Learning
FairLMs: A Turnkey Library for Fairness in Language Models
Jiale Zhang, Michael Larionov, Zichong Wang et al. (+2 more)
Fairness research on
language models involves measuring bias, applying mitigation methods, and examining the evidence on which an evaluation rests. Existing tools offer complementary functionality through different interfaces, so combining them requires reconciling model interfaces, evidence formats, access constraints, and result types before applicability can be checked or methods compared. We introduce FairLMs, a Python library that connects these activities through explicit declarations of model capabilities and input requirements. It provides 33 intrinsic and extrinsic metrics, 14 mitigation components spanning four intervention categories, 14
dataset and scoring-instrument diagnostics, adapters for the three
Transformer architectures and supported hosted completion APIs, and
benchmark loaders. Declarations are checked before execution and results carry the configuration under which they were obtained, so that compatible components can be combined, methods compared under a common protocol, and workflows extended to new models and
datasets. The source code is available at: https://github.com/FairLMs/FairLMs.
Read the full abstract ▾
All authors Jiale Zhang, Michael Larionov, Zichong Wang, Zhipeng Yin, Wenbin Zhang
Machine Learning
Multi-Subject Pretraining Enables Short-Calibration Personalization for Closed-Corpus Surface EMG Speech Decoding
Chenqian Le, Beatrice Fumagalli, Yasamin Esmaeili et al. (+5 more)
Surface electromyography (sEMG)-based silent
speech interfaces are limited by cross-user variability and calibration burden. We study a limited-data setting in which each of 27
speech-typical participants contributed less than 0.5 h of data (21.3 min on average) across Aloud and Mimed
speech. Within a closed 50-sentence corpus, we used leave-one-subject-out evaluation, initializing from a released single-subject checkpoint, pretraining on non-held-out participants, and fine-tuning on the target participant. This pipeline achieved 21.7% character error rate (CER) and 31.9% word error rate (WER), compared with 49.3% CER without target-subject calibration and 68.0% CER for direct checkpoint fine-tuning. Multi-subject pretraining from random initialization followed by fine-tuning reached 44.9% CER and did not converge under the fixed schedule in 5 of 27 folds, indicating substantial optimization and accuracy benefits from checkpoint initialization. Macro-averaged CER declined from 74.4% with one pretraining participant to 21.7% with 26. Three minutes of target-subject calibration achieved 20.5% CER and 31.7% WER, with no statistically significant difference from the full approximately 13-min pool (21.7% CER and 31.9% WER). A subject-specific adapter provided no detectable benefit. Excluding the five evaluation sentences from all sEMG model-training data increased CER and WER to 78.6% and 99.9%. These results support short-calibration personalization in a standardized-montage, closed-corpus setting.
Read the full abstract ▾
All authors Chenqian Le, Beatrice Fumagalli, Yasamin Esmaeili, Xupeng Chen, Tianyu He, Nikasadat Emami, Adeen Flinker, Yao Wang
Machine Learning · Cross-listed from cs.IR
Hybrid GPU-CPU Retrieval for Personalized Search at Ultra-Large Scale
Hao Fu, Jichao Sun, Baiting Zhu et al. (+15 more)
Embedding-based
retrieval on user-generated content at the trillion-document scale exposes a sharp conflict between two production demands: deep, expressive personalization for queries with rich user intent, and broad coverage of a massive inventory under fixed latency and resource budgets. We characterize this as the personalization-scale paradox: hosting the full serving inventory in GPU
memory is too resource intensive, while CPU compute cannot execute the same interaction-heavy model on the latency-critical path. We present a hybrid GPU-CPU co-serving system that resolves the paradox through orchestration rather than a new model class. A high-depth GPU pathway fuses
retrieval and interaction pre-ranking over a curated online pool on the order of a billion documents, while a high-breadth CPU pathway searches an independently selected online inventory roughly twenty times larger with lightweight personalized scoring. Either or both pathways can run per request; candidates are deduplicated before shared downstream ranking. The system is deployed in production. A full-system A/B test against the legacy CPU-only configuration improves model-scored relevance and substantive engagement, while separate pathway experiments show positive value at their own deployment scopes.
Retrieval logs show that the pathways contribute structurally distinct candidates, production serving measurements characterize their latency, and a matched capacity plan quantifies the economic rationale for assigning modeling depth to GPUs and inventory breadth to CPUs. Together, these results validate a practical, independently evolvable depth-breadth architecture for ultra-large-scale personalized search.
Read the full abstract ▾
All authors Hao Fu, Jichao Sun, Baiting Zhu, Qiaoling Liu, Yan Shi, Cheng Lu, Liu Liu, Yubo Wang, Xin Yao, Xiangyu Niu, Xu Dong, Wenhan Lyu, Chiyao Shen, Yinjie Huang, Minglei Chen, Shuai Ding, Li Fan, Xiao Kong
Machine Learning
MIRCID: Inferred Hub-miRNAs Drive Cross-Task Improvements in Drug Mechanistic Modeling
Xin Cao, Yigang Chen, Jiatong Xu et al. (+11 more)
Drug mechanism-of-action (MoA) modeling commonly relies on perturbational transcriptomes, but matched microRNA (miRNA) measurements are often unavailable. Inferred regulatory features offer a scalable way to reuse these data. Here, we present MIRCID, a framework comparing gene expression with inferred transcription factor (TF) activity and miRNA expression across pathway classification and similarity-based MoA
retrieval. HubmiRNet infers 414 pan-cancer hub miRNAs (HubmiRs) from 977 L1000 landmark genes, achieving a Pearson correlation coefficient of 87.72%; its 1,298-output variant also outperformed SiCmiR on the full-miRNA task (71.21% versus 67.30%). In the evaluated comparisons, miRNA augmentation provided more consistent gains than TF activity. Generic
embedding controls showed model-dependent utility, while complementarity analyses identified a distinct, partially linearly recoverable representation that retained gene-derived structure. Illustrative rescue cases linked improved classification to biologically plausible miRNA patterns in samples with weak transcriptional signatures. These findings support inferred HubmiRs as a biologically informed recoding of transcriptomic data for perturbational drug modeling, while leaving recovery of measured perturbational miRNA responses to further validation.
Read the full abstract ▾
All authors Xin Cao, Yigang Chen, Jiatong Xu, Ziyue Zhang, Xiang Cheng, Shenyu Wang, Yangyi Zhang, Xiaoxuan Cai, Shidong Cui, Zihao Zhu, Xiang Ji, Hsi-Yuan Huang, Yang-Chi-Dung Lin, Hsien-Da Huang
Machine Learning · Cross-listed from cs.CL
How Many Humans Are 32 LLM Judges Worth?
Chao Li, Yingying Yu and Yunfeng Li
A panel's human-equivalent size is target-specific. Matching a fixed 32-judge panel to empirical human label distributions on three ChaosNLI tasks yields two distinct effective sizes: distributional-error matching gives $\nu_{\mathrm{MSE}}=2.304$, $3.750$, and $3.445$, whereas spectral matching gives $\nu_H=4.242$, $6.459$, and $6.499$, a gap of $1.72$–$1.89\times$; a binary-error diagnostic credits the same panels with only $1.971$–$2.227$ effective votes. Extrapolating the distributional-error curve at fixed squared mean residual, mean member variance, and normalized mean covariance gives asymptotes of $2.392$, $3.990$, and $3.655$, with 32 judges already reaching $94.0$–$96.3\%$. An exact spectral identity explains the gap: error depends on member energy and on the orientation of residual variation relative to averaging, information that the participation ratio (PR) discards. A realizable hard-label construction confirms that higher spectral diversity can coexist with worse distribution recovery even under equal member energies and nonnegative correlations, and the consensus direction retains $\gamma_{\mathrm{co}}=43.8\%$, $33.7\%$, and $35.9\%$ of centered residual variance. An external check on CC-1000, a 1,000-item Civil Comments subset with a different panel, gives $\nu_H=2.84$. For panel choice, we establish an existence result and one feasible path: exhaustive enumeration at $k\in\{5,7\}$ shows that panels beating the accuracy-top-$k$ baseline on both accuracy and $\nu_H$ always exist, and greedily swapping at most two members reaches $24.8$–$56.0\%$ higher $\nu_H$ at $0.10$–$1.10$ percentage points higher accuracy. Our
dataset and code are available at https://github.com/Chao1208/32judges-votes.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.AR
Programming AMD XDNA NPUs with Open-source Compiler Tools: A FlashAttention Case Study
Erwei Wang, Ephrem Wu, Victor J. B. Jung et al. (+4 more)
Spatial NPUs such as AMD XDNA place compute tiles beside small local
memories and leave data movement between them to software. Mapping a multi-stage workload onto such a device is largely a question of where the intermediate tensors live. We report what we learned making those choices for FlashAttention with the open-source IRON and MLIR-AIR flows. We compare four reference designs on XDNA 1 and XDNA 2: one runs each operator separately, two stream between operators on chip, and one fuses all three
attention stages into a single kernel. The fused kernel holds the $\boldsymbol{QK}^{\mathsf T}$ scores in compute-tile local
memory and reduces partial results over the cascade interconnect, so the scores never return to shared MemTile
memory. On XDNA 2, it reaches 3.62 TFLOP/s over complete end-to-end execution, twice the IRON design, with 5.3 to 7.2 times the energy efficiency of the integrated GPU on the same chip at 2K
tokens and above. It covers twelve LLM configurations, from BERT to DeepSeek, up to 128K
tokens. Roofline analysis at each
memory level explains this result and shows when to stop. XDNA 1 has lower ridge points, so streaming on chip already reaches the compute-bound regime: the same fusion that doubles throughput on XDNA 2 is nearly wasted on XDNA 1. Comparing a mapping's operational intensity against each level's ridge point predicts which case applies before writing any code. Fuse until the mapping clears that ridge point, then stop. We release the reference designs as maintained open source.
Read the full abstract ▾
All authors Erwei Wang, Ephrem Wu, Victor J. B. Jung, Jiajie Li, Andre Rosti, Joseph Melber, Samuel Bayliss
Machine Learning · Cross-listed from cs.IR
Verify, Don't Trust: Agentic Model Development for Video Discovery Retrieval at Scale
Hao Fu, Baiting Zhu, Minglei Chen et al. (+2 more)
Large
language model (LLM)
agents can propose, implement, and evaluate model changes. Autoresearch loops demonstrate this capability through minutes-scale iterations on a self-contained program. Online autoresearch instead spans asynchronous systems, hours-long variants, and weeks-long campaigns that can influence a product. A completed run can still support an invalid conclusion when a code change is a no-op, data windows leak, evaluator semantics drift, or the two arms traverse different serving funnels. We present EvoPilot, a human-gated method for long-horizon online autoresearch. Role-specific
agents execute each round through a versioned domain skill and typed adapter. Durable records preserve experiments and failures; deterministic checks enforce recorded lessons. We study a 37-day campaign for the
retrieval system that powers
Video Deep Dive (VDD), an online experience for discovering follow-on
videos after a user opens a seed
video. The campaign covered seven directions and used an hourly refreshed index of hundreds of millions of
videos. Earlier manual experiments had not established a benefit from an interaction head. A primitive autoresearch attempt revisited the direction but incorrectly attributed an offline hit-rate decline of 22 percentage points to the head. We then introduced EvoPilot. Its human-gated verification traced the drop to a pre-existing evaluation defect that produced output depths of 3,000 and 600. After repair, a matched comparison measured an offline improvement of 3.20 percentage points. Post-study replay and mutation tests rejected invalid comparisons while admitting valid counterparts. Durable state recovered an interrupted round, and artifact reuse avoided approximately five GPU-hours. Separately, a seven-day randomized online evaluation estimated a 0.66% relative increase in the VDD slice of Good Search Result Rate for Retention (GSRR).
Read the full abstract ▾
All authors Hao Fu, Baiting Zhu, Minglei Chen, Yinjie Huang, Shuai Ding
Machine Learning · Cross-listed from quant-ph
From Trainability Diagnostics to Optimization Claims: Boundaries and Controls in Variational Quantum Optimization
Pilsung Kang
Barren plateau diagnostics characterize whether gradient signal remains available for training, but surviving signal need not translate into successful optimization. We study this trainability–optimization gap at the level of optimizer steps. Treating coefficient-weighted
Hamiltonian-term gradients as task-like components, we introduce step-level diagnostics and derive an exact bridge between signed termwise organization, directional activity, and first-order descent. Resolving this bridge into standard first-order geometry shows that the apparent organization–activity factors are not independent optimization axes and that, at fixed state and update norm, the raw gradient maximizes first-order descent of the summed objective. We compare vanilla gradient descent, a deterministic
Hamiltonian-term PCGrad variant, and probe-gated LSO-PCGrad on transverse-field Ising model instances with hardware-efficient and
Hamiltonian variational ansatzes, together with matched controls for update norm and probe budget. Blind projection can improve an organization diagnostic while worsening final energy and first-order predictability. After conditioning on standard first-order geometry, residual term-space composition shows no reproducible material incremental association with realized descent, while optimizer-relative update norm shows positive material associations in some settings without cross-regime reproducibility. Matched controls provide no resolved final-energy benefit attributable to the projected direction, and the improvement of LSO-PCGrad is more consistent with probe-based search and step-norm adaptation than with
Hamiltonian-term projection itself. These results show that gradient-structure diagnostics can characterize trainability and update geometry without serving as standalone evidence of optimization benefit, which requires controls matched on update norm and search budget.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.CV
Multiclass Semantic Segmentation of Wildland Fire Images Using Context-Aware Centralized Copy-Paste Data Augmentation
Joon Tai Kim, Nishanth Kunchala, Vishv Patel et al. (+5 more)
Producing accurate annotations for deep learning based image
segmentation is both costly and labor intensive. This challenge is especially evident in wildland fire applications, where accurately labeled
datasets are scarce due to the difficulty of collecting and annotating dynamic fire scenes. To address this problem, our previous work introduced the Centralized Copy-Paste Data Augmentation (CCPDA) method for semantic
segmentation of wildland fire imagery, which generates artificial training samples by randomly pasting fire clusters from source images onto target images. However, random placement can produce contextually unrealistic scenes, such as fire burning on asphalt. In this paper, we present a context-aware strategy designed specifically to improve data quality and realism in small multiclass wildland fire
datasets, ensuring that augmented samples remain contextually meaningful. The proposed method restricts fire placement to semantically valid target regions and selects the location whose Ash-Vegetation composition most closely matches the source context. This approach preserves existing fire regions in the target image, prevents unrealistic placements, and maintains contextual accuracy by generating images that resemble real wildland fire scenes. We evaluate the Context-Aware CCPDA strategy through numerical analysis and comparisons with other augmentation methods by a weighted sum-based multi-objective optimization (MOO) approach. The results confirm that the context-aware data augmentation strategy leads to improved
segmentation performance and contextual realism, outperforming other augmentation procedures.
Read the full abstract ▾
All authors Joon Tai Kim, Nishanth Kunchala, Vishv Patel, Tianle Chen, Ziyu Dong, Daniel Ospina Acero, Roger Williams, Mrinal Kumar
Machine Learning · Cross-listed from cs.RO
Beiming Li, Jaime Romero, Jonathan Diller et al. (+2 more)
Learned
navigation policies typically consume observations as a temporally ordered history, with positional encodings tying each observation to when it was seen, making it difficult to reuse experience from earlier traversals of an environment. Systems that do reuse such experience usually construct an explicit representation, such as a map or a topological graph, and plan on it. We propose VNT-PA (Visual
Navigation Transformer with Pose
Attention), a
transformer planner whose context is a set of depth keyframes indexed by camera pose. With camera poses as positional encoding,
attention depends on the pose differences between keyframes rather than on their temporal order. VNT-PA is trained to imitate a shortest-path planner operating on the ground-truth scene mesh, predicting actions by querying the spatial context with only its current pose and the goal position. On point-goal
navigation in HM3D validation scenes, VNT-PA reaches 93.3% success and 90.4% success weighted by path length (SPL), outperforming baselines that encode the same context as a temporal sequence or treat pose as an input feature, in both
navigation performance and training efficiency. Because the spatial context is a pose-indexed set, frames from different trajectories can be fused at test time. The planner also degrades more gracefully under localization noise than a conventional baseline which plans on explicit maps. These results show that pose-stamped experience can serve directly as the environment representation for a learned planner, and that making
attention depend on pose differences, rather than on temporal order, speeds up training and improves long-horizon
navigation.
Read the full abstract ▾
All authors Beiming Li, Jaime Romero, Jonathan Diller, Vijay Kumar, Alejandro Ribeiro
Machine Learning
Reliability-Centered Evaluation of Sparse Longitudinal CT Lesion-Size Forecasting with Conformal Interval Calibration and Gompertz-Inspired Regularization
Lingfei Kong
Sparse longitudinal CT follow-up limits lesion-size
forecasting when only a few prior observations are available. We constructed a five-visit DLT-derived same-lesion trajectory
benchmark from DeepLesion and Deep Lesion Tracker (DLT), yielding 205 trajectories from 129 patients. We compared an exploratory conventional sparse-to-final analysis with a primary fixed visit-index horizon design predicting the common log change from T3 to T4 while progressively adding earlier observations, evaluating predictive accuracy, uncertainty reliability, post-hoc conformal interval calibration, subgroup performance, and Gompertz-inspired trajectory regularization. The evaluated methods showed partially overlapping point-prediction accuracy but distinct uncertainty behavior. Mean held-out RMSE across ten training seeds was 0.4726, 0.4305, 0.4499, and 0.4513 for m = 1, 2, 3, 4, indicating the lowest mean RMSE at m = 2; additional history did not improve RMSE. At m = 4, raw Cohort-Level Feature GP coverage was near the 95% nominal level, whereas MC Dropout, Deep Ensemble, and residual-scale intervals were conservative. Patient-level conformal calibration generally produced near-nominal or conservative coverage at the cost of wider intervals. Patient-grouped development cross-validation selected lambda* = 0 for the Gompertz-inspired term. A global population reference frequently opposed lesion-level change directions, and prediction difficulty varied across anatomical subgroups. Overall, additional historical observations provided limited predictive benefit once the prediction horizon was controlled, while predictive accuracy, uncertainty reliability, and trajectory consistency did not necessarily improve together, and should be evaluated jointly in sparse longitudinal imaging.
Read the full abstract ▾
Machine Learning
SWE-Proof: Can Language Models Resolve Real-World Issues with Machine-Checked Proofs?
George Ma, Benjamin Mikek, Haoyu Li et al. (+9 more)
Ensuring the correctness of LLM-generated code is a core challenge for modern software engineering.
Benchmarks for agentic code generation check correctness with held-out test suites, which are inherently incomplete and increasingly susceptible to memorization. Formal verification avoids both problems, but existing work covers only standalone tasks whose specifications are given as input, not real issues, which touch large repositories and state intent in vague natural
language. We present Benchproofer, a pipeline that turns a coding task with a known correct patch into a formally verified one: it writes a specification for the new code, summarizes the existing functions that code calls with axioms, and admits an instance only after mechanical and
adversarial gates agree. Applying it to SWE-bench Verified yields SWE-Proof, 500 real issues whose correctness is formally verified rather than tested, and it extends to SWE-bench Pro. Evaluating Claude Opus 4.8, we find that verification catches what tests miss: a quarter of test-passing patches admit counterexamples, which a structured natural-
language specification does not fix, while a correct formal one lifts resolution from 85% to 95%. Writing that specification is the hard part: an
agent that must write its own gains nothing over an unaided baseline, and only 56% of those specifications pass our audit. The usual failure is faithfulness, a specification that constrains part of the required behavior and leaves the rest free. Specification quality still tracks the outcome, failing on 92% of unresolved instances against 51% of resolved ones, making faithful specification synthesis a concrete open problem.
Read the full abstract ▾
All authors George Ma, Benjamin Mikek, Haoyu Li, Ferhat Erata, Yuhao Zhang, Zeren Shui, Behrooz Omidvar Tehrani, Jun Huan, Murali Krishna Ramanathan, Somayeh Sojoudi, Hao Zhou, Anoop Deoras
Machine Learning · Cross-listed from cs.AI
Implicit Rule Induction with Test-Time Task Embeddings in ARC-like Tasks
Adrien Deliège, Claas Beger, Marc Van Droogenbroeck et al. (+1 more)
The Abstraction and
Reasoning Corpus and related
benchmarks evaluate whether AI models can solve novel
reasoning tasks, but often leave unclear whether success reflects inference of the intended underlying rule or reliance on shortcuts. We address this gap by studying test-time task
embeddings in
Vision ARC (VARC), a model in which a pre-trained backbone is complemented by a trainable
embedding representing the transformation rule. In the original VARC, test-time training (TTT) is jointly applied to the backbone and task
embedding. Here we introduce a novel two-step TTT protocol: first finetune only the task
embedding (Embed-TTT), then freeze it and finetune the backbone. Across ARC-AGI-1, ConceptARC, and two controlled
datasets with known rules, Embed-TTT consistently yields improved task
embeddings, ones that align better with underlying task rules, improve
embedding-based
retrieval, and enable accurate linear probing of known rules. Qualitatively, Embed-TTT identifies more semantically meaningful relations between test and train tasks on ARC-AGI-1. We also show that optimizing only task
embeddings (less than 0.01% of model parameters) already solves a non-trivial fraction of ARC-AGI-1, ConceptARC, and Mini-ARC tasks, while the full two-step pipeline improves final performance. Finally, we show that Embed-TTT recovers the underlying
geometric structure of parametric rules and learns compositional capabilities that enable rule-wise interpolation, but not extrapolation. These findings support a clearer separation between rule induction and rule execution in ARC-like evaluations, motivating
benchmarks that better distinguish in-distribution from out-of-distribution rules.
Read the full abstract ▾
All authors Adrien Deliège, Claas Beger, Marc Van Droogenbroeck, Melanie Mitchell
Machine Learning
TierKV: Long-Context On-Device LLMs via Predictive Multi-Tier KV Caching
Zhihao Shu, Md Musfiqur Rahman Sanim, Jie Hu et al. (+4 more)
Large
language models (LLMs) are moving onto mobile devices for increasingly diverse workloads over text, images,
video, and audio. These applications often require long contexts, making the Key-Value (KV) cache a dominant
memory bottleneck because it grows linearly with sequence length and is accessed at every decoding step. Prior work reduces KV-cache footprint through low-rank compression,
token eviction, or flash offloading, but the resulting reconstruction overhead, irreversible
token loss, or I/O stalls can offset the benefit of saving
memory. We present TierKV, a mobile LLM inference framework built on Predictive Multi-Tier Cache Optimization (PMCO). Before decoding starts, PMCO predicts future cache demand from prefill hidden states and jointly assigns
tokens to exact, low-rank, and flash-offloaded tiers under the device
memory and accuracy budgets. This formulation retains access to the full context, removes the circular dependency of reactive eviction, and admits a closed-form solver that selects tier boundaries and per-layer ranks at runtime. Across eight text,
vision, and audio models on three mobile SoCs, TierKV improves prefill throughput by up to 17.6x over existing mobile LLM frameworks, reduces RAM-resident KV cache by 12.5-34%, thereby enabling substantially longer contexts under the same
memory budget, while incurring only minor accuracy degradation.
Read the full abstract ▾
All authors Zhihao Shu, Md Musfiqur Rahman Sanim, Jie Hu, Kun Yuan, Minghai Qin, Gagan Agrawal, Wei Niu
Machine Learning
M2G-LLM: Enhancing Clinical Prediction via Multimodal Graph Reasoning and LLM Context Injection
Inyoung Choi, Sukwon Yun, Jiayi Xin et al. (+3 more)
Integrating diverse data modalities — such as clinical notes, laboratory results, and medical imaging — is essential for advancing clinical decision-making. While Large
Language Models (LLMs) have shown remarkable performance in processing unstructured clinical text, their limited capacity to incorporate non-text modalities hinders their broader utility in healthcare applications. Here, we introduce M2G-LLM (
Multimodal MedGraph-LLM), a novel framework that enhances LLMs with
multimodal integration and
alignment via Graph Neural Networks (GNNs). Our approach models temporal relationships between patient visits, propagates information across clinically similar patients, and aligns heterogeneous data sources to construct enriched
multimodal context vectors. These vectors are injected into the intermediate layers of the LLM, enabling joint
reasoning over textual and non-textual modalities. We evaluate M2G-LLM on the MIMIC-IV and MIMIC-CXR
datasets, demonstrating improvements in clinical prediction tasks over strong baseline models. Our results highlight the promise of combining the
language understanding of LLMs with the relational
reasoning capabilities of GNNs for comprehensive,
multimodal healthcare analysis.
Read the full abstract ▾
All authors Inyoung Choi, Sukwon Yun, Jiayi Xin, Jie Peng, Tianlong Chen, Qi Long