Pith. sign in

REVIEW 1 major objections 60 cited by

A simple convolutional architecture outperforms LSTMs on diverse sequence tasks while showing longer effective memory.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.3

2026-05-11 19:30 UTC pith:SVE6YRE4

load-bearing objection This paper's abstract makes the case that a basic convolutional model beats LSTMs on many standard sequence benchmarks and handles longer memory, backed by released code. the 1 major comments →

arxiv 1803.01271 v2 pith:SVE6YRE4 submitted 2018-03-04 cs.LG cs.AIcs.CL

An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling

Shaojie Bai , J. Zico Kolter , Vladlen Koltun This is my paper
classification cs.LG cs.AIcs.CL
keywords sequence modelingconvolutional networksrecurrent networksLSTMbenchmark evaluationtemporal modelingeffective memory
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper conducts a systematic comparison of generic convolutional and recurrent networks on standard sequence modeling benchmarks. It finds that a basic convolutional model surpasses canonical recurrent networks such as LSTMs in performance across the tested tasks. The convolutional approach also maintains longer effective memory without relying on recurrence. This evidence questions the default link between sequence modeling and recurrent architectures. The authors therefore position convolutional networks as a natural starting point for new sequence tasks.

Core claim

The authors evaluate generic convolutional and recurrent architectures across a broad range of standard sequence modeling tasks used to benchmark recurrent networks. Their results indicate that a simple convolutional architecture outperforms canonical recurrent networks such as LSTMs, while demonstrating longer effective memory. They conclude that the common association between sequence modeling and recurrent networks should be reconsidered, and that convolutional networks should be regarded as a natural starting point for sequence modeling tasks.

What carries the argument

A simple convolutional architecture that processes sequences in parallel and captures long-range dependencies without recurrence.

Load-bearing premise

The chosen tasks and datasets represent general sequence modeling challenges and both architectures are implemented and compared fairly without hidden advantages.

What would settle it

A standard sequence modeling benchmark where an LSTM records higher accuracy or F1 score than the convolutional model, or where the convolutional model's effective memory span proves shorter than the LSTM's.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

1 major / 0 minor

Summary. The manuscript conducts a systematic empirical evaluation of generic convolutional and recurrent architectures for sequence modeling on standard benchmark tasks. It claims that a simple convolutional architecture outperforms canonical recurrent networks such as LSTMs across a diverse range of tasks and datasets while demonstrating longer effective memory, and concludes that convolutional networks should be regarded as a natural starting point for sequence modeling tasks. Code is made available at a GitHub repository.

Significance. If the reported results hold, the work would be significant in challenging the default association of sequence modeling with recurrent networks and in providing evidence for convolutional architectures as a competitive or superior alternative, supported by the public code release which enables reproducibility and independent verification.

major comments (1)
  1. [Abstract] Abstract: The central claims of outperformance over LSTMs and longer effective memory are presented as direct results without any quantitative metrics, tables, task descriptions, implementation details, hyperparameter protocols, or statistical tests in the provided manuscript; these elements are load-bearing for assessing whether the data support the conclusion that convolutional networks should replace recurrent ones as the default.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their review and for highlighting the need for clarity on how the abstract relates to the supporting evidence in the manuscript. We address the major comment below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claims of outperformance over LSTMs and longer effective memory are presented as direct results without any quantitative metrics, tables, task descriptions, implementation details, hyperparameter protocols, or statistical tests in the provided manuscript; these elements are load-bearing for assessing whether the data support the conclusion that convolutional networks should replace recurrent ones as the default.

    Authors: The abstract is a concise summary of the paper's primary conclusions, following standard academic conventions that reserve detailed evidence for the main text. The full manuscript supplies all requested elements: task and dataset descriptions appear in Section 4, the experimental protocol including implementation details and hyperparameter search is given in Section 5, quantitative results with direct LSTM comparisons are reported in Tables 1--3 and Figures 2--4, and the longer effective memory analysis is presented in Section 5.3 with supporting dilation and receptive-field experiments. Performance differences are shown consistently across multiple independent runs and diverse benchmarks, supporting the claim that the convolutional architecture is a competitive starting point. The public code release enables independent verification of all reported numbers. revision: no

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper is a purely empirical study that performs a systematic comparison of convolutional and recurrent architectures on standard sequence modeling benchmarks. No derivations, first-principles results, fitted parameters presented as predictions, or self-referential equations appear in the abstract or described methodology. Claims rest on direct experimental measurements against external tasks and datasets, with code released for reproducibility. No load-bearing steps reduce to inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no free parameters, axioms, or invented entities are described beyond standard neural network components and benchmark tasks.

pith-pipeline@v0.9.0 · 5423 in / 1056 out tokens · 69730 ms · 2026-05-11T19:30:04.145631+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling." pith.science (2026). https://pith.science/paper/SVE6YRE4

@misc{pith2026180301271,
  author       = {Pith},
  title        = {Pith review of: An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SVE6YRE4}},
  note         = {Machine review of arXiv:1803.01271}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

For most deep learning practitioners, sequence modeling is synonymous with recurrent networks. Yet recent results indicate that convolutional architectures can outperform recurrent networks on tasks such as audio synthesis and machine translation. Given a new sequence modeling task or dataset, which architecture should one use? We conduct a systematic evaluation of generic convolutional and recurrent architectures for sequence modeling. The models are evaluated across a broad range of standard tasks that are commonly used to benchmark recurrent networks. Our results indicate that a simple convolutional architecture outperforms canonical recurrent networks such as LSTMs across a diverse range of tasks and datasets, while demonstrating longer effective memory. We conclude that the common association between sequence modeling and recurrent networks should be reconsidered, and convolutional networks should be regarded as a natural starting point for sequence modeling tasks. To assist related work, we have made code available at http://github.com/locuslab/TCN .

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 60 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. BAH Dataset for Ambivalence/Hesitancy Recognition in Videos for Digital Behavioural Change

    cs.CV 2025-05 accept novelty 8.0

    Introduces the BAH dataset with 1,427 annotated videos for multimodal recognition of ambivalence/hesitancy in digital behavior change contexts.

  2. Efficiently Modeling Long Sequences with Structured State Spaces

    cs.LG 2021-10 unverdicted novelty 8.0

    S4 is an efficient state space sequence model that captures long-range dependencies via structured parameterization of the SSM, achieving state-of-the-art results on the Long Range Arena and other benchmarks while bei...

  3. Twoblock clustering trees with coskewness-based dimension reduction: recovering piecewise multivariate linear regimes

    stat.ME 2026-07 conditional novelty 7.0

    A coskewness-maximizing split objective in a twoblock regression tree recovers piecewise linear regimes better than covariance splits and matches black-box ensembles on two multivariate benchmarks.

  4. ConTex: Reformulating Counterfactual Generation For Time Series Forecasting

    cs.LG 2026-06 unverdicted novelty 7.0

    ConTex learns a global intervention strategy via a decomposed temporal-conditional encoder architecture to generate consistent, sparse counterfactuals for time series models in a single forward pass.

  5. OpenGlass: Ultra-Low-Power On-Device AI Eyewear with Event-based Vision

    cs.CV 2026-06 unverdicted novelty 7.0

    OpenGlass is an open-source smart glasses platform using event-based vision and event-driven power management to achieve 11.5 hours of continuous on-device ML on a 200 mAh battery, demonstrated with 83.94% cross-subje...

  6. Scale-Equivariant Generative Forecasting: Weight-Tied Dilated Convolutions, Wavelet Scattering Inputs, and Spectral-Consistency Training for Self-Similar Time Series

    cs.LG 2026-05 unverdicted novelty 7.0

    Presents SE-WaveNet with weight-tied dilated convolutions plus wavelet and spectral components that reproduces empirical scaling collapse on financial returns while using L times fewer convolutional parameters.

  7. U-STS-LLM A Unified Spatio-Temporal Steered Large Language Model for Traffic Prediction and Imputation

    cs.LG 2026-05 unverdicted novelty 7.0

    U-STS-LLM uses a spatio-temporally steered LLM with dynamic attention bias generation to achieve state-of-the-art results on long-horizon traffic forecasting and high-missing-rate imputation while remaining parameter-...

  8. Clin-JEPA: A Multi-Phase Co-Training Framework for Joint-Embedding Predictive Pretraining on EHR Patient Trajectories

    cs.LG 2026-05 unverdicted novelty 7.0

    Clin-JEPA supplies a multi-phase co-training method for JEPA pretraining on EHR trajectories that achieves converging latent rollouts and improved multi-task AUROC on MIMIC-IV data.

  9. TailedTS: Benchmark Dataset for Heavy-Tailed Time Series Prediction and Periodicity Quantification

    cs.LG 2026-05 unverdicted novelty 7.0

    TailedTS supplies 24.69 billion Wikipedia page-view records as a public benchmark for heavy-tailed time series forecasting and periodicity analysis, revealing weaker periodic structure in high-traffic pages.

  10. SIGMA-ASL: Sensor-Integrated Multimodal Dataset for Sign Language Recognition

    cs.HC 2026-05 unverdicted novelty 7.0

    SIGMA-ASL is a multimodal dataset with 93,545 word-level ASL clips from Kinect RGB-D, mmWave radar, and dual IMUs, plus benchmarking protocols for single- and multi-modal recognition.

  11. AegisTS: A Hierarchical Agent System with Reinforcement Learning for Multivariate Time Series Data Cleaning

    cs.DB 2026-05 unverdicted novelty 7.0

    AegisTS uses a two-level RL agent architecture with a dual-stage reward to jointly optimize cleaning order and method selection for multivariate time series, delivering up to 96% better cleaning quality and 27% better...

  12. BadmintonGRF: A Multimodal Dataset and Benchmark for Markerless Ground Reaction Force Estimation in Badminton

    cs.CV 2026-05 unverdicted novelty 7.0

    BadmintonGRF is a new public multimodal dataset and benchmark that pairs multi-view video with instrumented GRF for markerless load estimation in badminton.

  13. GAFSV-Net: A Vision Framework for Online Signature Verification

    cs.CV 2026-04 unverdicted novelty 7.0

    GAFSV-Net encodes online signatures as asymmetric Gramian Angular Field images and processes them with dual-branch ConvNeXt plus cross-attention to outperform sequence-based baselines on DeepSignDB and BiosecurID.

  14. Autocorrelation Reintroduces Spectral Bias in KANs for Time Series Forecasting

    cs.LG 2026-04 unverdicted novelty 7.0

    Temporal autocorrelation reintroduces spectral bias in KANs for time series forecasting, which DCT preprocessing can mitigate.

  15. A Convolutional Neural Network-Derived Catalog of Solar Flares from Soft X-Ray Observations

    astro-ph.SR 2026-04 unverdicted novelty 7.0

    The CNN-derived catalog detects over seven times more solar flares than the GOES catalog and extends the power-law distribution of flare peak fluxes to smaller sizes.

  16. Adversarial Robustness of Deep State Space Models for Forecasting

    cs.LG 2026-04 conditional novelty 7.0

    Spacetime SSM forecasters represent optimal Kalman predictors for autoregressive data but remain vulnerable to model-free attacks that exploit local linearity and increase error by over 33% compared to projected gradi...

  17. CalM: A Self-Supervised Foundation Model for Population Dynamics in Calcium Imaging Data

    q-bio.QM 2026-04 unverdicted novelty 7.0

    CalM uses a discrete tokenizer and dual-axis autoregressive transformer pretrained self-supervised on calcium traces to outperform specialized baselines on population dynamics forecasting and adapt to superior behavio...

  18. T1: One-to-One Channel-Head Binding for Multivariate Time-Series Imputation

    cs.LG 2026-02 conditional novelty 7.0

    T1 uses one-to-one channel-head binding in a CNN-Transformer hybrid to achieve robust multivariate time-series imputation, cutting average MSE by 46% versus the next-best baseline across 11 datasets even at 70% missingness.

  19. MELT: A Behavioral Trace Dataset for High-Risk Memecoin Launch Detection

    cs.CR 2026-02 unverdicted novelty 7.0

    MELT is the first behavioral trace dataset for high-risk memecoin launch detection on Solana, providing 122 features, risk annotations, and ML benchmarks that reduce investment loss when used for selection.

  20. Universal Approximation Theorems for Dynamical Systems with Infinite-Time Horizon Guarantees

    math.DS 2026-02 conditional novelty 7.0

    Neural ODEs can approximate Morse-Smale and continuous-attractor dynamical systems over infinite time in an ε-δ sense, provided limit-cycle periods are matched exactly.

  21. Causal Time Series Generation via Diffusion Models

    cs.LG 2025-09 unverdicted novelty 7.0

    CaTSG is a unified diffusion model for causal time series generation that handles observational, interventional, and counterfactual tasks via backdoor adjustment and abduction-action-prediction.

  22. Sundial: A Family of Highly Capable Time Series Foundation Models

    cs.LG 2025-02 conditional novelty 7.0

    Sundial uses TimeFlow Loss for native pre-training of Transformers on continuous time series from TimeBench, achieving SOTA point and probabilistic forecasting with millisecond inference.

  23. What Causes Performance Degradation in Cross-Subject EEG Classification?

    cs.CE 2024-10 unverdicted novelty 7.0

    Controlled experiments attribute cross-subject EEG classification degradation to inter-subject variability in multi-class tasks and shortcut learning in single-class tasks.

  24. Deep Time Series Models: A Comprehensive Survey and Benchmark

    cs.LG 2024-07 unverdicted novelty 7.0

    This survey and benchmark of deep time series models using the released TSLib library finds that models with specific structures perform well only on distinct analysis tasks.

  25. Time-LLM: Time Series Forecasting by Reprogramming Large Language Models

    cs.LG 2023-10 conditional novelty 7.0

    Time-LLM reprograms frozen LLMs for time series forecasting via text prototypes and Prompt-as-Prefix, outperforming specialized models in standard, few-shot, and zero-shot settings.

  26. Convolutional and Deep Learning based techniques for Time Series Ordinal Classification

    cs.LG 2023-06 unverdicted novelty 7.0

    First benchmarking of ordinal adaptations of CNN and DL methods for time series shows they outperform nominal TSC techniques on ordinal metrics across 29 selected problems.

  27. A Time Series is Worth 64 Words: Long-term Forecasting with Transformers

    cs.LG 2022-11 conditional novelty 7.0

    PatchTST uses subseries patching and channel-independent Transformers to deliver significantly better long-term multivariate time series forecasting and strong self-supervised transfer performance.

  28. Tropical time series, iterated-sums signatures and quasisymmetric functions

    math.RA 2020-09 unverdicted novelty 7.0

    Defines iterated-sums signatures over commutative semirings (tropical case emphasized) for time-series feature extraction and links them to quasisymmetric functions over semirings.

  29. The Importance of Encoder Choice:A Tabular-Image Study

    cs.LG 2026-07 conditional novelty 6.5

    Tabular encoder choice reorders multimodal rankings, can erase apparent fusion gains, and requires non-vanilla extraction for in-context learning models to avoid train-test representation shift.

  30. U3DWind: A Low Altitude Wind Field Dataset and Benchmark for Urban Air Mobility

    physics.flu-dyn 2026-07 conditional novelty 6.5

    A multi-city LBM-LES urban wind dataset (720 cases, 10 m, five Chinese megacities) plus five UAM-oriented benchmarks for prediction, reconstruction, risk, siting, and noise.

  31. Learning What to Remember: Test-Time Training via Context Distillation

    cs.CL 2026-08 conditional novelty 6.0

    IP-TTCD distills the hidden-state gap between a long-window teacher and a short-window student into MLP fast weights, improving long-context language modeling and retrieval over DeltaNet, Gated DeltaNet, SWA, and IP-TTT.

  32. Versatile On-device Adaptation at the Edge by Unifying Few-shot, Zero-shot, Continual, and In-context Learning

    cs.LG 2026-07 conditional novelty 6.0

    ECL unifies few-shot, continual, zero-shot, and in-context learning on a single TCN-based edge chip, with first hardware baselines on several tasks.

  33. MATERO-RCA: Mode-Aware Trajectory-Level Energy-Based Root-Set Optimization for Industrial Root Cause Analysis

    eess.SP 2026-07 conditional novelty 6.0

    MATERO-RCA jointly optimizes root set, root-effect modes, and repaired counterfactual trajectories, with a MILP-accelerated best-bound search that reports strong benchmark RCA results.

  34. StatePlay: State-Aware Game World Models for Mechanics-Consistent Generation

    cs.CV 2026-07 conditional novelty 6.0

    Coupling explicit state prediction with MoT-style video generation raises mechanics fidelity of Street Fighter 3 rollouts by about 18.6% over stateless game world models.

  35. Every Client Is an Environment: Federated De-confounding for Spatio-Temporal Forecasting

    cs.LG 2026-07 conditional novelty 6.0

    Federated aggregation of aligned environmental prototype codebooks acts as implicit back-door adjustment, with de-confounding error bounded linearly by average client confounding strength.

  36. Label-free Industrial Fault Detection via Adversarial Inverse Reinforcement Learning: A System for Run-to-Failure Prognostics

    cs.LG 2026-07 reject novelty 6.0

    Adversarial inverse reinforcement learning over consecutive vibration samples yields label-free early fault warnings, but the headline robustness claim depends on a threshold fitted to the test set.

  37. T-STAR: A Large-Scale Benchmark for Spatio-Temporal Panoptic Scene Graph Generation in Satellite Video

    cs.CV 2026-07 conditional novelty 6.0

    T-STAR provides 1.19M instance masks and 3.83M spatio-temporal relationship triplets from 150 Jilin-1 satellite videos, defining a new task of panoptic scene graph generation in satellite video.

  38. Deep learning-based prediction of time-resolved adhesive forces in viscoelastic Hertzian contacts

    cs.LG 2026-07 conditional novelty 6.0

    A Tabor-conditioned LSTM predicts complete time-resolved adhesive force trajectories of viscoelastic Hertzian contacts from displacement history alone, with median pull-off error of about 2.2% and hysteresis error of ...

  39. PeakFlow: Peak-Guided Coarse-to-Refined Modeling for EEG-Based Dynamic Affective Trajectory Prediction

    cs.HC 2026-07 conditional novelty 6.0

    PeakFlow improves EEG-based dynamic emotion trajectory prediction by adding a peak-guided residual refiner that suppresses terminal-peak bias, cutting the false-terminal peak rate from 62.6% to 5.4% on SEED-VII.

  40. Fourier Geometric Wind Power Forecasting with Numerical Weather Prediction

    cs.AI 2026-07 conditional novelty 6.0

    A multimodal model coupling SCADA measurements with numerical weather prediction via rotation-invariant geometric features and Fourier neural operator fusion outperforms more than 20 baselines on three UK wind farms.

  41. Knowledge-Assisted Multi-Graph Dependency Learning for Multivariate Time Series Anomaly Detection in Multi-Stage Industrial Processes

    cs.LG 2026-07 conditional novelty 6.0

    Incorporating sensor-group and process-flow knowledge into graph construction improves multivariate time-series anomaly detection on multi-stage industrial processes.

  42. Real-Time Detection of Charge Jumps in Superconducting Qubits with a Convolutional Neural Network

    quant-ph 2026-07 conditional novelty 6.0

    A dilated causal CNN quantized to fixed point and synthesized to an FPGA detects charge jumps in superconducting qubits at 6.19 μs latency with 0.843 efficiency, close to the 0.866 of the offline χ2 method on |Δq|∈[0.1,0.5]e.

  43. Catching Disguised Transients with ASTRANet: Anomaly-Aware Spectroscopic Classification and Conformal Calibration

    astro-ph.IM 2026-07 conditional novelty 6.0

    ASTRANet combines a redshift-free spectral classifier, a 16-score anomaly detector, and conformal prediction to identify and calibrate uncertainty for out-of-taxonomy astronomical transients.

  44. KinEMbed: Decoding Kinematics from Electromyography via Cross-Modal Contrastive Learning

    cs.LG 2026-07 conditional novelty 6.0

    Cross-modal InfoNCE alignment of EMG features to joint-angle geometry yields modestly better continuous hand kinematics regression than standard representation baselines on held-out NinaPro DB8 sessions.

  45. EgoInertia-MI: A Multimodal Egocentric Vision and IMU Benchmark for Motor Impairment Assessment

    cs.CV 2026-07 conditional novelty 6.0

    A synchronized egocentric-video + multi-IMU benchmark of 19 simulated motor-impairment activities yields multimodal Macro-F1 of ~0.78 for severity and ~0.93 for action recognition.

  46. Learning a Normal World Model for Few-Shot Boundary-Calibrated Abnormality Detection

    cs.LG 2026-06 unverdicted novelty 6.0

    A hypergraph entropic normal-world model learns normality from abundant normal sensor data and calibrates abnormality boundaries with few examples, reporting AUROC 0.9983 on the most complex NASA C-MAPSS subset.

  47. INDEQS: Informed Neural controlled Differential EQuationS

    cs.LG 2026-06 unverdicted novelty 6.0

    INDEQS is a graph-informed NCDE variant that separates inner hidden-state mixing from outer vector-field mixing and reports lower MAE than uninformed NCDEs on synthetic advection data and real river/traffic tasks when...

  48. SenFlow: Inter-Sentence Flow Modeling for AI-Generated Text Detection in Hybrid Documents

    cs.CL 2026-06 unverdicted novelty 6.0

    SenFlow integrates graph-based inter-sentence propagation with linear-chain CRF decoding for sentence-level AI-generated text detection and reports SOTA results on the new MOSAIC benchmark of 16k hybrid documents.

  49. TS-Fault: Benchmarking Time Series Forecasters Against Structural Faults

    cs.LG 2026-06 unverdicted novelty 6.0

    TS-Fault benchmark finds clean-data accuracy anti-correlates with robustness to structural faults, with all catastrophic failures under mechanism-level faults and foundation models most fragile.

  50. From Coarse to Fine: Managing Temporal Granularity in Spatio-Temporal Data for Fine-Grained Traffic Prediction

    cs.AI 2026-06 unverdicted novelty 6.0

    STRP is a granularity-aware model that predicts fine-grained spatio-temporal traffic from coarse inputs via tree convolution and inverse dilated convolution, outperforming baselines on six datasets in window-based and...

  51. End-to-End Control of a Powered Knee-Ankle Prosthesis Towards Unified, Tuning-Free Assistance

    cs.RO 2026-06 conditional novelty 6.0

    An end-to-end TCN controller trained on multi-terrain data from 18 transfemoral amputees was deployed in real time on four users and reproduced key torque and angle scalings across five locomotion modes without subjec...

  52. Second-Order Path Kernel Interpolation Formulas in Machine Learning

    cs.LG 2026-06 unverdicted novelty 6.0

    Derives second-order path-kernel interpolation formulas for gradient descent, SGD, and momentum training, adding curvature terms and a concentration estimate around the expected prediction.

  53. Pretraining Recurrent Networks without Recurrence

    cs.LG 2026-06 unverdicted novelty 6.0

    SMT reduces RNN training to supervised learning on memory transitions (m_t, x_{t+1}) to m_{t+1} obtained from a Transformer encoder, enabling time-parallel training with O(1) gradient paths.

  54. Dynamic Short Convolutions Improve Transformers

    cs.LG 2026-06 unverdicted novelty 6.0

    Dynamic short convolutions applied to key/query/value projections and linear layers in Transformers yield consistent performance gains and 1.33-1.60x compute advantages over standard models on language modeling from 1...

  55. Trajectory-Aware Clinical Risk Prediction via Severity-Grounded Knowledge Graphs and Retrieval-Augmented Generation

    cs.AI 2026-06 conditional novelty 6.0

    A severity-aware knowledge-graph retrieval-augmented LLM pipeline reports large gains in mortality and readmission prediction on MIMIC-III/IV.

  56. Dynamics Are Learned, Not Told: Semi-Supervised Discovery of Latent Dynamics Geometries For Zero-Shot Policy Adaptation

    cs.RO 2026-06 unverdicted novelty 6.0

    Contrastive learning bounds the Lipschitz constant of a trajectory dynamics encoder to support outcome-centric zero-shot adaptation in MuJoCo robotics tasks under severe dynamics shifts.

  57. Beyond Holistic Models: Systematic Component-level Benchmarking of Deep Multivariate Time-Series Forecasting

    cs.LG 2026-05 unverdicted novelty 6.0

    TSCOMP is the first large-scale benchmark that deconstructs deep multivariate time series forecasters into fine-grained components, builds a corpus of over 20,000 evaluations, and shows that corpus-driven component se...

  58. PHGNet: Prototype-Guided Hypergraph Construction for Heterogeneous Spatiotemporal Forecasting

    cs.AI 2026-05 unverdicted novelty 6.0

    PHGNet proposes prototype-guided hypergraph construction plus global-local representations and temporal attention to model high-order spatiotemporal dependencies in traffic data and reports better performance than pri...

  59. ReactiveGWM: Steering NPC in Reactive Game World Models

    cs.CV 2026-05 unverdicted novelty 6.0

    ReactiveGWM introduces a decoupled diffusion architecture for player-NPC interactions that learns game-agnostic response logic for zero-shot strategy transfer across games.

  60. Clin-JEPA: A Multi-Phase Co-Training Framework for Joint-Embedding Predictive Pretraining on EHR Patient Trajectories

    cs.LG 2026-05 unverdicted novelty 6.0

    A five-phase co-training framework enables stable JEPA pretraining on EHR trajectories, producing converging latent rollouts and higher multi-task AUROC than baselines on MIMIC-IV ICU data.