finding
pending-review
finding:probe-based-ranking-reduces-harmful-behavior-by-63-via-datapoint-filtering

Probe-based ranking reduces harmful behavior by 63% via datapoint filtering

xiao-aranguri-probe-data-attribution-2026.md
Frontmatter (11 fields)
{
  "doc": "xiao-aranguri-probe-data-attribution-2026.md",
  "author": null,
  "context": "Primary quantitative result: probe method outperforms gradient-based and LLM-judge alternatives at lower computational cost.",
  "enrichment": {
    "is_stale": true
  },
  "norm_label": "Probe-based ranking reduces harmful behavior by 63% via datapoint filtering",
  "source_url": null,
  "graphify_id": "probe_ranking_reduction",
  "source_file": "xiao-aranguri-probe-data-attribution-2026.md",
  "imported_from": "/Users/antonborzov/Documents/Research.nosync/papers/extract_typed_out/xiao-aranguri-probe-data-attribution-2026/graph.json",
  "extracted_type": "finding",
  "graphify_file_type": "finding"
}

Outgoing (3)

answered_by (1)

Incoming (0)

None.

Mentions (1)

  • papers-typed
    xiao-aranguri-probe-data-attribution-2026.md