Abstract
Techniques are described for mode-aware ranking in recommendation systems. Session-level behavioral signals are aggregated over a sliding window to form a behavioral feature vector including at least dwell time, scroll velocity, topic autocorrelation, and interaction depth ratio. A cognitive mode detector outputs a probability distribution over cognitive modes including browse, decide, and flow. A feature gating network uses the mode probability distribution to generate gate weights for feature groups, and the gate weights are applied to candidate item feature groups to produce gated features used for scoring and ranking. Within-session transitions may be managed using hysteresis and smooth interpolation of gate weights across a transition window. In agentic implementations, the detected mode selects and injects mode-specific scoring instructions into an LLM ranking prompt.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Anonymous, "Cognitive Mode Detection and Conditional Feature Gating for Content Ranking Systems", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/10752