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 modelCheck whether a 4-bit quantized encoder still matches its bf16 embeddings before reporting a similarity
- Kind
- SkillSkill
- Pricing
- Open sourceOpen source
- Platforms
- Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf, GooseClaude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, Windsurf, Goose
- GitHub stars
- 41.3K41.3K
- Last commit
- Nov 2026Nov 2026
- License
- MITMIT
- 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 projectCheck 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.
