Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces SIFT (Self-Improving, Frozen-gate Training), a dynamic document classification system that addresses the challenges of labeling and retraining in enterprise settings. By utilizing a low-cost SPLADE sparse encoder and a LightGBM head, SIFT escalates only low-confidence classifications to an LLM judge, which then updates the labeled corpus based on real production data. The system's innovative frozen-gate mechanism ensures safety by preventing regressions during autonomous retraining, allowing for continuous improvement without extensive human intervention.
SIFT transforms document classification by enabling a self-improving system that reduces labeling costs to near zero while maintaining safety and accuracy.
Document classification is a solved problem in the laboratory and an unsolved one in the enterprise. The blocker is rarely model architecture; it is the labeling project that must precede a model and the institutional fear of letting a model retrain itself once one exists. We present SIFT (Self-Improving, Frozen-gate Training), a dynamic classifier service, which attacks both. SIFT serves classification from a deliberately cheap, CPU-bound pipeline, a SPLADE sparse encoder feeding a LightGBM head, and escalates only the low-confidence minority of pages to an LLM judge. The judge's verdicts are written back into a labeled corpus, so the expensive model continuously teaches the cheap one: the escalation rate falls, the corpus grows from production traffic rather than from an up-front annotation effort, and accuracy compounds with use. Onboarding a new document family requires only a declarative bundle, label space, anchor phrases, and a judge glossary, not a labeling project. The harder problem is safety: an autonomously retraining classifier can silently regress. SIFT resolves this with a two-part promote gate, a critical-label F1 regression check plus a frozen golden regression set the model is never trained on, either of which vetoes promotion. This turns"retrain monthly without a human"from reckless into routine. We describe the architecture, the self-feeding corpus loop, the frozen-gate promotion mechanism, and an illustrative multi-domain deployment, and we discuss the economics of a classifier whose marginal labeling cost trends toward zero.