Skip to content

Datanalysis credit risk vs Embedding anisotropy check

Datanalysis credit risk is an open-source agent skill for analytics and data. Embedding anisotropy check is an open-source agent skill for knowledge and memory.

  • Datanalysis credit risk is best for clean a raw credit dataset before training a pre-loan risk model.
  • Embedding anisotropy check is best for check whether a 4-bit quantized encoder still matches its bf16 embeddings before reporting a similarity.
What it is
Datanalysis credit risk is an open-source agent skill for analytics and data.
Embedding anisotropy check is an open-source agent skill for knowledge and memory.
Best for
Clean a raw credit dataset before training a pre-loan risk model
Check whether a 4-bit quantized encoder still matches its bf16 embeddings before reporting a similarity
Kind
Skill
Skill
Pricing
Open source
Open source
Platforms
Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf, Goose
Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf, Goose
GitHub stars
41.3K
41.3K
Last commit
Nov 2026
Nov 2026
License
MIT
MIT
Open source
Good for
Clean a raw credit dataset before training a pre-loan risk model; Find variables with unstable PSI across months for a loan portfolio; Drop features with high missing rates and weak IV from a scorecard project
Check whether a 4-bit quantized encoder still matches its bf16 embeddings before reporting a similarity; Verify that two adjacent hidden-state layers encode the same information in a probe comparison; Evaluate a retrieval index's query and passage pairs against an unrelated-pair baseline

What Datanalysis credit risk does

  • Data loading and formatting

    Loads raw credit data and standardizes it using configured date, label, organization and key columns.

  • Abnormal month filtering

    Removes months with too few bad samples or too few total samples.

  • Missing rate screening

    Calculates overall and per-organization missing rates and drops features above a threshold.

  • IV and PSI variable filtering

    Drops features with weak information value or unstable population stability across months and organizations.

  • Null Importance and correlation pruning

    Removes noise features by label permutation and trims highly correlated variables by gain ranking.

  • Excel cleaning report

    Exports a workbook with per-step details, statistics and distributions.

What Embedding anisotropy check does

  • Per-side anisotropy measurement

    Reports how strongly the vectors on each side share a common direction, using mean cosine between different rows.

  • Floor-relative matched scoring

    Compares matched pair cosines with unrelated pair cosines from the same comparison, before and after centring.

  • Retrieval rank check

    Reports how often the true partner ranks first, alongside its chance rate of 1/n.

  • Mean source guidance

    Explains when to fit centring means on a separate train split versus the compared rows themselves.

  • Reporting rules for gaps

    Advises on averaging signed differences, giving spread and detectable effect size, and keeping floors with results.

  • Bundled Python script

    Provides a command-line script that takes paired NumPy arrays and prints the diagnostic figures.

Datanalysis credit risk vs Embedding anisotropy check: questions

Is Datanalysis credit risk or Embedding anisotropy check cheaper?

Both are open source. Compare current plans on each official site.

What is the difference between Datanalysis credit risk and Embedding anisotropy check?

Datanalysis credit risk is an open-source agent skill for analytics and data. Embedding anisotropy check is an open-source agent skill for knowledge and memory.

Keep comparing