SQuAT: Spherical Lattice Vector Quantization-Aware Training
2 judges. Scored by two of the three judges. The third reads only the day’s strongest candidates. How far stories are checked
SQuAT is a quantization-aware training method for low-bit large language models that splits each weight into a direction and a magnitude, quantizing directions with an E8-derived spherical codebook and gains with scalar quantizers.
Score
0.56total under rubric v2.0
- Significance0.57
- Reader interest0.61
- Audience breadth0.42
- Practical relevance0.57
- Substantiated0.57
- Technical depth0.79
- New information0.87
- Counts against: Promotional0.06
- Counts against: Overclaim0.34
Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).
| Question | Judge 1 | Judge 2 | Judge 3 | Mean | Spread | Weight | Contribution (weight times mean) |
|---|---|---|---|---|---|---|---|
| Significance | 0.65 | 0.50 | — | 0.57 | 0.15 | +0.25 | +0.143 |
| Reader interest | 0.62 | 0.60 | — | 0.61 | 0.02 | +0.20 | +0.122 |
| Audience breadth | 0.44 | 0.40 | — | 0.42 | 0.04 | +0.10 | +0.042 |
| Practical relevance | 0.59 | 0.55 | — | 0.57 | 0.04 | +0.15 | +0.085 |
| Substantiated | 0.54 | 0.60 | — | 0.57 | 0.06 | +0.10 | +0.057 |
| Technical depth | 0.78 | 0.80 | — | 0.79 | 0.02 | +0.10 | +0.079 |
| New information | 0.94 | 0.80 | — | 0.87 | 0.14 | +0.10 | +0.087 |
| Promotional | 0.06 | 0.05 | — | 0.06 | 0.01 | −0.15 | −0.008 |
| Overclaim | 0.33 | 0.35 | — | 0.34 | 0.02 | −0.15 | −0.051 |
| Sum of contributions | +0.557 | ||||||
| Claim-check penalty | 0.000 | ||||||
| Total, as published | 0.557 | ||||||
Spread is the highest minus the lowest judge answer. marks a spread above 0.40: the judges disagree, and the story is marked contested. Judge 3 did not score this story; its mean uses the other two.
Also covered by
No other coverage was clustered with this story.
Elsewhere
Elsewhere: HN —
Recorded for context only, checked 24 Oct 2026, 08:54 UTC. It played no part in the ranking.