I built an epistemic gate to stop LLM data poisoning during fine-tuning

说明与用途

Beatriz 是一种轻量级代理，用于防止在 LLM 微调期间的定向数据中毒。它使用余弦相似度与不可变语料库进行验证，并采用组合损失函数，在多域基准上实现 93% 的精确度和 80% 的召回率。

问题

Malicious data poisoning during fine‑tuning can silently corrupt specific facts while aggregate metrics improve, posing a stealthy threat to small teams and startups fine‑tuning local models.

结果

Across five architectures (GPT‑2, Qwen‑2.5‑0.5B, TinyLlama‑1.1B, Pythia‑1.4B, Phi‑3‑mini‑4k‑instruct) the gate achieved a robust truth margin (+4.19 ± 0.08) with 0.93 precision and 0.80 recall on a 30‑fact benchmark. Ablation showed that filtering alone provides 65 % of the benefit, with contrastive loss adding the remaining 35 %.

工作流

• 1. Generate candidate text from a generative source. 2. Pass the text through the Beatriz Gate. 3. Compute cosine similarity against an immutable anchor corpus. 4. Apply a composite loss (cross‑entropy + contrastive Softplus) to enforce truthfulness. 5. Forward the filtered text to the student model for fine‑tuning.
• 6. Evaluate the fine‑tuned model on a held‑out multi‑domain benchmark to confirm robustness.

证据

Beatriz acts as a lightweight, non‑invasive defensive proxy (~0.1 ms decision latency) that sits between a generative source and a training student model.

来源

Key Results (EXP08–EXP16): Multi‑architecture validation: Tested across GPT‑2 (124M), Qwen‑2.5‑0.5B, TinyLlama‑1.1B, Pythia‑1.4B, and Phi‑3‑mini‑4k‑instruct (3.8B).

来源

Scaled Held‑Out (EXP16): Validated on a rigorous 30 multi‑domain unattacked facts benchmark, consolidating a robust truth margin of +4.19 ± 0.08 with 0.93 Precision and 0.80 Recall on the gate.

来源