Surprised by Attention: Predictable Query Dynamics for Time Series Anomaly Detection | Signal Canvas | ScienceToStartup

← Back to Paper

Surprised by Attention: Predictable Query Dynamics for Time Series Anomaly Detection

Stale81d agoVerification pending / evidence receipt incomplete

Clone Repo Export Brief Open in Build Loop Connect with Author

Use This Via API or MCP

Use this Signal Canvas via API or MCP

Route this paper proof surface into REST, MCP, or developer workflows while preserving the same evidence receipt and related-resource context.

Signal Canvas guide REST guide MCP guide

Page Freshness

Signal Canvas proof surface

Canonical route: /signal-canvas/surprised-by-attention-predictable-query-dynamics-for-time-series-anomaly-detection

stale

Proof freshness: stale
Proof status: unverified
Display score: 8/10
Last proof check: 2026-03-19
Score updated: 2026-04-02
Score fresh until: 2026-05-02
References: 0
Source count: 0
Coverage: 50%

This page is showing the last landed evidence receipt and score bundle because the latest proof data is outside the freshness window.

Agent Handoff

Surprised by Attention: Predictable Query Dynamics for Time Series Anomaly Detection

Canonical ID surprised-by-attention-predictable-query-dynamics-for-time-series-anomaly-detection | Route /signal-canvas/surprised-by-attention-predictable-query-dynamics-for-time-series-anomaly-detection

REST example

curl https://sciencetostartup.com/api/v1/agent-handoff/signal-canvas/surprised-by-attention-predictable-query-dynamics-for-time-series-anomaly-detection

MCP example

{
  "tool": "search_signal_canvas",
  "arguments": {
    "mode": "paper",
    "paper_ref": "surprised-by-attention-predictable-query-dynamics-for-time-series-anomaly-detection",
    "query_text": "Summarize Surprised by Attention: Predictable Query Dynamics for Time Series Anomaly Detection"
  }
}

source_context

{
  "surface": "signal_canvas",
  "mode": "paper",
  "query": "Surprised by Attention: Predictable Query Dynamics for Time Series Anomaly Detection",
  "normalized_query": "2603.12916",
  "route": "/signal-canvas/surprised-by-attention-predictable-query-dynamics-for-time-series-anomaly-detection",
  "paper_ref": "surprised-by-attention-predictable-query-dynamics-for-time-series-anomaly-detection",
  "topic_slug": null,
  "benchmark_ref": null,
  "dataset_ref": null
}

Paper mode· single-doc scopescope: surprised-by-attention-predictable-query-dynamics-for-time-series-anomaly-detection

Preparing verified analysis

GitHub Code Pulse

Stars

1

Health

B

Last commit

6/8/2026

Forks

0

Open repository

Claim map

Strong 8Mixed 0Weak 0

Evidencepartial
AxonAD improves ranking quality and temporal localization over strong baselines.
Implicationpartial
Directly stated in abstract with specific dataset and metric details
Verificationpartial
partial
Evidencepartial
Multivariate time series anomalies often manifest as shifts in cross-channel dependencies rather than simple amplitude excursions.
Implicationpartial
Directly stated as foundational premise in abstract
Verificationpartial
partial
Evidencepartial
Residual-based detectors can miss such anomalies when flexible sequence models still reconstruct signals plausibly despite altered coordination.
Implicationpartial
Directly stated limitation of existing methods in abstract
Verificationpartial
partial
Evidencepartial
We introduce AxonAD, an unsupervised detector that treats multi-head attention query evolution as a short horizon predictable process.
Implicationpartial
Directly stated core methodological innovation in abstract
Verificationpartial
partial
Evidencepartial
Ablations confirm that query prediction and combined scoring are the primary drivers of the observed gains.
Implicationpartial
Directly stated in abstract from ablation studies
Verificationpartial
partial
Evidencepartial
At inference, reconstruction error is combined with a tail-aggregated query mismatch score, which measures cosine deviation between predicted and target queries on recent timesteps.
Implicationpartial
Directly stated methodological approach in abstract
Verificationpartial
partial
Evidencepartial
This dual approach provides sensitivity to structural dependency shifts while retaining amplitude-level detection.
Implicationpartial
Directly stated benefit of the method in abstract
Verificationpartial
partial
Evidencepartial
On proprietary in-vehicle telemetry with interval annotations
Implicationpartial
Directly stated evaluation domain with specific annotation type
Verificationpartial
partial