AIAny
Icon for item

Knowledge-Geometry Decoupling: Refreshable Pretrained Transfer for Streaming Recommendation

A pretrain-then-transfer method for streaming recommendation that decouples refreshable behavioral knowledge from task-specific geometry to enable continual model refresh without downstream interference; introduces Behavioral Multi-Token Prediction and Anchored Calibration Residual and shows 4–12% offline gains plus live Shopee A/B lifts.

Introduction

Behavioral distributions in industrial recommenders drift continually, so a static pretrained model either becomes stale or its continual refresh breaks downstream task adaptation. The core insight here is that what a recommender should learn (transferable behavioral knowledge) is different from how a specific task arranges representations (task geometry). Separating those concerns lets you refresh the learned behavior signal continuously while keeping downstream task-specific structure stable.

Key Findings
  • Behavioral Multi-Token Prediction (BMTP): only use future items that are collaboratively or semantically related as prediction targets, reducing spurious transitions in next-token training — so what: yields cleaner, more transferable behavioral priors for downstream fine-tuning.
  • Knowledge-Geometry Decoupling (KGD) architecture: a refreshable encoder stores pretrained behavioral knowledge, a read-only cross-attention task learner consumes encoder states, and task-specific geometry is written via Anchored Calibration Residual (ACR) orthogonal to pretrained embeddings — so what: prevents task gradients from corrupting pretrained knowledge and allows independent refresh of the encoder.
  • Robust empirical gains: KGD improves 4–12% over strong pretrain-transfer baselines across eight public benchmarks and sustains advantage over a 90-day production stream where baselines stagnate — so what: demonstrates both offline and long-term online robustness.
  • Production impact: fully deployed at Shopee; a live A/B on Homepage Search increased GMV per user by 1.75% and advertising revenue by 1.53% — so what: the method translates to measurable business value in a large-scale setting.
Who it's for and tradeoffs

Great fit if you operate or research large-scale recommenders with continual user-behavior drift and need a pretrain-transfer workflow that can be refreshed online without invalidating downstream models. It is particularly relevant for teams that already use large pretrained encoders or need stable downstream task adapters in production. Look elsewhere if your setting is small-scale (no continual drift), if you cannot afford separate parameter stores for encoder vs. task learner, or if tight latency/parameter constraints forbid read-only cross-attention and residual calibration layers.

Mechanism overview

BMTP tightens supervision by selecting semantically/collaboratively related future items rather than raw next-item tokens. KGD then enforces ownership: the encoder is the sole owner of behavioral knowledge and is refreshable; the task learner accesses encoder outputs but writes geometry changes via ACR, which is constrained to be orthogonal to pretrained embeddings to avoid interference. This simple ownership separation is the practical lever that enables continual refresh without downstream degradation.

Information

More Items

Autonomously proposes, modifies, executes, and evaluates ML experiments to study recursive self-improvement in machine learning engineering. Implements an open stack (OpenMLE-Gym, -RL, -Evo) and post-trains Frontis-MA1 (35B) around four evolution operators (Draft, Improve, Debug, Crossover); releases model weights and the full codebase.

Provides a portable C++ inference runtime to deploy embodied AI models (vision–language–action and world–action) on heterogeneous robot hardware, enabling latency-first batch-1 closed-loop control. Key features include modular multi-rate layers, fused low-latency inference, and extensible head/IO plugins.

Standardizes representation-level evaluation for tabular encoders by exporting row-, column-, and table-level embeddings and probing them with shared lightweight heads across three suites (TRL-CTbench, TRL-Rbench, TRL-DLTE). Supplies curated benchmark assets and task rewrites (50 OpenML tables, 123 targets, a 47,772-table DLTE lake) to enable fair cross-paradigm comparison.