03 · Confidence · Composition layer

Lattice

Confident recommendations

A patent-pending gating layer on your sequential backbone — blend in extra structure when the session has support; fall back to your model when it does not.

Move your cursor — confidence crystallizes near the probe.

RecSys 2026 — under reviewPatent pendingEvaluation bundle ↗arXiv preprint ↗
+31.7%Gated HR@10 vs LSTM backbone (MovieLens, 30 paired seeds)
OFF ≡ backboneGate off: predictions match your backbone exactly (we verify parity)
ON ≡ fusionGate on: same behavior as always-on fusion on supported rows

Not a new recommender — a gate on yours

Lattice sits on the backbone you already ship. A binary ON/OFF switch decides whether an auxiliary scoring path gets blended in — not whether items disappear from the list.

1

Your backbone

LSTM, transformer, SASRec — any sequential ranker you already train and serve. Lattice never replaces it.

2

Confidence gate

When confidence is below threshold, the gate stays OFF. Rankings match your backbone exactly — auditable, not quietly down-weighted.

3

Backbone-native fusion

When ON, archetypes fit in your backbone's embedding space and fusion kicks in. Always-on fusion scores higher on pooled benchmarks; gating trades peak HR for control you can explain.

Validated results

MovieLens primary eval: 30 paired seeds, full-catalog ranking, same test rows every seed. Lifts are vs each backbone alone — not vs published SOTA. Want to reproduce independently? LorianBannis@banlys.com.

BackboneDatasetGated lift (HR@10)Ungated liftSeedsNote
LSTMMovieLens+31.7%+58.7%30Headline result
TransformerMovieLens+13.3%+20.9%30Backbone-native archetypes
SASRecMovieLens+17.0%+26.9%30Contextual baseline + Lattice
LSTMAmazon Electronics+10.7%+17.3%30Supporting — 5k test rows

Headline: LSTM +31.7% gated (+58.7% ungated). Transformer +13.3% and SASRec +17.0% use backbone-native archetypes. Amazon +10.7% is supporting cross-domain evidence (30 seeds, 5k test rows). We are not selling peak benchmark mode — we sell an ON/OFF switch that does no harm when off.

When Lattice composes vs defers

GATE ON

Compose

  • Session in support for learned behavioral structure
  • Auxiliary fusion path enabled on top of your backbone
  • On gate-ON rows, matches always-on fusion in our eval
  • Primary lift: +31.7% HR@10 vs LSTM on MovieLens (30 seeds)
GATE OFF

Defer

  • Low support, short history, or unfamiliar input
  • Auxiliary path fully disabled — not down-weighted
  • Predictions match your backbone exactly (parity verified)
  • Under distribution shift, gate tends to stay off — metrics match baseline

The trade-off (stated plainly)

Always-on fusion scores higher on pooled benchmarks in our ablation (+58.7% vs +31.7% gated on MovieLens). We do not sell peak benchmark mode. We sell a switch you can audit: off means your backbone; on means full fusion gain on supported rows.

Baseline vs gated vs always-on

Lattice decides whether an auxiliary scoring path gets composed with your backbone — not whether items vanish from the list. Gate OFF: rankings match your backbone exactly. Gate ON: same as always-on fusion on that session.

Session support (demo)55%
Gate threshold θ40%

GATE ON — auxiliary fusion active

Gated ranking matches always-on fusion on this session (same behavior as our MovieLens eval: OFF ≡ LSTM, ON ≡ ungated fusion).

Backbone only (baseline)

  1. 1.Viral TikTok Hit(Comedy)
  2. 2.Trending Sequel(Action)
  3. 3.Hype Train Movie(Drama)
  4. 4.Inception(Sci-Fi)
  5. 5.The Dark Knight(Action)

Lattice (gated)

  1. 1.Inception(Sci-Fi)
  2. 2.The Dark Knight(Action)
  3. 3.Interstellar(Sci-Fi)
  4. 4.The Matrix(Sci-Fi)
  5. 5.Pulp Fiction(Crime)

Always-on fusion (ungated)

  1. 1.Inception(Sci-Fi)
  2. 2.The Dark Knight(Action)
  3. 3.Interstellar(Sci-Fi)
  4. 4.The Matrix(Sci-Fi)
  5. 5.Pulp Fiction(Crime)

Mock rankings for education only — composition policy (OFF ≡ backbone, ON ≡ fusion), not item filtering. Does not expose production confidence formulas or training pipelines.

Research & commercial pilots

Pilots focus on three things: coverage when ON, lift when ON, and zero harm when OFF. Hosted API access comes after a pilot validates — not self-serve yet.

A Banlys product · Patent pending · RecSys 2026 under review · LorianBannis@banlys.com