Illustrative demoSimulates confidence-gating behavior for education — not connected to production inference.

Illustrative demo

Explore how binary gating changes recommendations. For research inquiries, get in touch.

Gated vs Ungated

A simplified example: popularity-based lists often surface low-confidence picks. Gating withholds recommendations the system does not trust.

Lattice blocks

3 risky recommendations

that traditional recsys would show anyway

Viral TikTok Hit, Trending Sequel, Hype Train Movie

Threshold65%

Traditional (Top-5 by popularity)

Always show top 5 by engagement — no confidence check.

1. Viral TikTok Hit(Comedy)
38% confrisky
2. Inception(Sci-Fi)
94% conf
3. Trending Sequel(Action)
45% confrisky
4. The Dark Knight(Action)
91% conf
5. Hype Train Movie(Drama)
52% confrisky

Lattice (confidence-gated)

Only recommend when confidence ≥ 65%.

1. Inception(Sci-Fi)
94%
2. The Dark Knight(Action)
91%
3. Interstellar(Sci-Fi)
88%
4. Pulp Fiction(Crime)
85%
5. Fight Club(Drama)
82%
6. The Matrix(Sci-Fi)
79%
7. Parasite(Thriller)
76%

Traditional: 5 items (avg 64% confidence)

Lattice: 7 items (avg 85% confidence — higher quality)