Gradian: Open-Source Tool to Pinpoint Poison Data Causing LLM Fine-Tuning Failures
vylara-ai · reddit · 2026-08-03
To solve the common pain point of models degrading silently during fine-tuning, a developer built Gradian, an open-source debugging tool.
Gradian computes per-example gradients over the LoRA adapter to rank and cluster training data, identifying exactly which examples erode specific capabilities. It also catches silent configuration errors like sequence truncation and missing EOS tokens. The author shared a counterintuitive finding: perfectly formatted poisoned data can be misjudged as helpful by standard gradients, requiring the subtraction of actual output gradients for accurate detection.
More from coding & agent
- Open-Source Multi-Agent Framework: Recursive Exploration & Synthesis — tom_doerr · 2026-08-03
- Open-Sourced Director/Implementer Delegation Protocol for Coding Agents — produckyou · 2026-08-03
- Meta Open-Sources Internal Design System Used Across 13K+ Apps — Vjeux · 2026-08-03
- Migrating from GPT-4o to GPT-5.1: Handling RAG Agent Response Style Regressions — IncreaseLocal2574 · 2026-08-03
- Using MCP to Seamlessly Invoke GlobalGPT in Codex for Writing and Video Generation — thetripathi58 · 2026-08-03
- Open-Source Skill Watches Videos and Generates Structured Notes — Roger_M_Taylor · 2026-08-03