Does weight-space programming need to touch the hidden layers?
Comments on “Does weight-space programming need to touch the hidden layers?”
3 judges. Scored by all three judges. Only the day’s top stories get the claim check. How far stories are checked
An output-head experiment on Qwen3.5-0.8B edited only three output rows, leaving every transformer tensor unchanged, and scored 55.2% on MATH-500 against 52.2% for a published Astra-edited reference run that changed internal feed-forward layers.
Why it ranked
No. 122 of 122 with a score of 0.56. The panel rated it highest on technical depth (0.91) and new information (0.82), lowest on audience breadth (0.45). The judges disagreed most on substantiated (spread 0.42), so the story is marked contested. Placed No. 122, No. 105 by score alone: the 44th AI & machine learning story of the day (−0.080), and close to “Editing one row of an LLM's output layer to stop it overthinking”, No. 20 on Monday 16 November 2026 (−0.033). The placement terms move stories; they are not part of the score.
- AI judgeJudge 2claude-haiku-5-5
- Technical LLM interpretability experiment with real benchmark numbers and frank caveats on failures and noise; narrow audience, modest significance, detailed methodology but limited independent replication.
- AI judgeJudge 3gpt-5.6-luna
- A highly reproducible, technically detailed experiment showing output-head weight edits can alter reasoning termination, with benchmark results, caveats, artifacts, and failure analysis. Its importance is meaningful but limited by confounded comparisons and incomplete controls.
Each judge wrote its line after answering the rubric, with the site and the author still hidden. Every answer behind this is in the table under Score.
Score
0.56total under rubric v2.3Contestedjudges differ by 0.42 on substantiated
- Significance0.49
- Reader interest0.67
- Audience breadth0.45
- Practical relevance0.55
- Substantiated0.82
- Technical depth0.91
- New information0.82
- Counts against: Promotional0.06
- Counts against: Overclaim0.22
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) |
|---|---|---|---|---|---|---|---|
| SignificanceHow much does this change what people in tech know or can do? | 0.48 | 0.30 | 0.68 | 0.49 | 0.38 | +0.24 | +0.117 |
| Reader interestWould a well-informed senior engineer or tech founder want to read this today, and likely pass it on? | 0.55 | 0.55 | 0.90 | 0.67 | 0.35 | +0.26 | +0.173 |
| Audience breadthHow many people who build technology does it concern? | 0.48 | 0.30 | 0.58 | 0.45 | 0.28 | +0.13 | +0.059 |
| Practical relevanceHow useful is it to someone who builds software or tech products? | 0.54 | 0.40 | 0.72 | 0.55 | 0.32 | +0.22 | +0.122 |
| SubstantiatedHow well is it backed? Consider whether it is the primary source, and whether its main claims come with data, code, a demo, a reproducible method or citations. | 0.97 | 0.55 | 0.93 | 0.82 | 0.42 , above the contested threshold | +0.05 | +0.041 |
| Technical depthHow technically substantive is it? | 1.00 | 0.80 | 0.94 | 0.91 | 0.20 | +0.05 | +0.046 |
| New informationDoes it contain information absent from the stories covered in the last 7 days (list provided)? | 0.80 | 0.70 | 0.96 | 0.82 | 0.26 | +0.05 | +0.041 |
| PromotionalIs it mainly marketing with little substance (a press release, ad or sponsored post with no real information)? A launch post that carries real news is not promotional. | 0.06 | 0.10 | 0.01 | 0.06 | 0.09 | −0.19 | −0.011 |
| OverclaimDo the headline or conclusions claim more than the body supports? | 0.28 | 0.35 | 0.04 | 0.22 | 0.31 | −0.11 | −0.025 |
| Sum of contributions | +0.563 | ||||||
| Claim-check penalty | 0.000 | ||||||
| Total, as published | 0.564 | ||||||
Spread is the highest minus the lowest judge answer. marks a spread above 0.40: the judges disagree, and the story is marked contested.
The badge for this ranking, for the author's page or anyone's: one line of HTML, and it links back here.
<a href="https://rubric.hrih.si/editions/2026-11-17/122/"><img src="https://rubric.hrih.si/badge/2026-11-17/122.svg" alt="Ranked #122 on Rubric, 17 Nov 2026" width="252" height="28"></a>Also covered by
No other coverage was clustered with this story.
Elsewhere
Elsewhere: HN 3 pts, not on front page
Recorded for context only, checked 17 Nov 2026, 05:47 UTC. It played no part in the ranking.