SFT Leaves a Reproducible Signature in Activation Geometry: Calibrating Post-Training Interpretability Against Run-to-Run Noise

arXiv:2610.17093v1 Announce Type: new Abstract: A model diff credits post-training with whatever separates a base checkpoint from one post-trained descendant, presuming a rerun would land close by. On the sparse co-activation graph of Qwen3-1.7B it does not: two LoRA-SFT reruns sit over half as far apart as either sits from the base. We score each base-to-post contrast against the run-to-run band of its own recipe, and show that k runs certify no better than p=1/(k+1) under run exchangeability, however large the effect, so the band z-score is a descriptive scale. SFT leaves a large, reproducible signature: base o SFT displaces the graph by 17.8 pairwise-distance SDs beyond the mean distance between four independent LoRA-SFT runs, and the pre-registered estimator replicates on Llama-3.1-8B with Llama-Scope. At 29 steps each, four SFT and four DPO runs occupy disjoint displacement ranges, a gap 5.0imes and 5.6imes the spreads of the two arms, without any band. All bands use graphs built on one GPU architecture. Three other readings stop short of a verdict: DPO band counts (3/4, 1/4, 3/4 contrasts outside at 12, 29, 60 steps) show no consistent dose ordering while its displacement stays flat; light GRPO is a secondary non-detection, not an absence, and we run no equivalence test; and the graph concentration read as an RL signature is already in the zero-RL base. A pre-registered causal sweep at three edge budgets on 4,816 held-out prompts found the displaced sources load-bearing (12imes a matched random control in mean, 3.5imes in median at 10,000 edges), but their stage-specificity interaction reversed at two budgets and the third failed its faithfulness gate, so we claim calibrated structural displacement, not a changed mechanism. Scope: one backbone, one SFT recipe and general-text probe states for the full protocol.

Source: arXiv cs.AI | 2026-10-15

Related