Edition of Tuesday 17 November 2026

Interpretability-guided weight edits for Gemma 270M, no training

Comments on “Interpretability-guided weight edits for Gemma 270M, no training”

3 judges. Scored by all three judges. Only the day’s top stories get the claim check. How far stories are checked

A Hacker News post by an individual developer describes making Google's 270-million-parameter Gemma 3 model better at following instructions without any gradient-based training, instead editing a few weights directly.

Read the original at balajsaleem.github.io. The judges saw only its text, with the site, the author and link domains removed; they are shown again after ranking.

Why it ranked

No. 102 of 122 with a score of 0.57. The panel rated it highest on new information (0.84) and technical depth (0.83), lowest on audience breadth (0.53). The judges disagreed most on significance (spread 0.42), so the story is marked contested. Placed No. 102, No. 93 by score alone: the 35th AI & machine learning story of the day (−0.080). The placement terms move stories; they are not part of the score.

AI judgeJudge 2claude-haiku-5-5
Interesting interpretability-guided weight-editing experiment on a small LLM with clear before/after results, but self-reported, single model, small per-category test sets, and not peer-reviewed; technically substantive and AI-focused.
AI judgeJudge 3gpt-5.6-luna
Technically detailed, potentially important evidence that interpretability-guided weight edits improve instruction following without retraining. Claims are supported by experiments and caveats, but missing numerical results, small test sets, limited reproducibility, and no peer review reduce confidence.

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

0.57total under rubric v2.3Contestedjudges differ by 0.42 on significance

Judges’ mean answer per question, on a scale from 0 to 1. Questions marked minus count against the score.
  • Significance0.56
  • Reader interest0.75
  • Audience breadth0.53
  • Practical relevance0.59
  • Substantiated0.59
  • Technical depth0.83
  • New information0.84
  • Counts against: Promotional0.11
  • Counts against: Overclaim0.45

Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).

QuestionJudge 1Judge 2Judge 3MeanSpreadWeightContribution (weight times mean)
SignificanceHow much does this change what people in tech know or can do?0.670.300.720.560.42 , above the contested threshold+0.24+0.135
Reader interestWould a well-informed senior engineer or tech founder want to read this today, and likely pass it on?0.750.600.900.750.30+0.26+0.195
Audience breadthHow many people who build technology does it concern?0.670.350.580.530.32+0.13+0.069
Practical relevanceHow useful is it to someone who builds software or tech products?0.660.400.720.590.32+0.22+0.130
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.0.700.450.630.590.25+0.05+0.030
Technical depthHow technically substantive is it?0.900.700.880.830.20+0.05+0.041
New informationDoes it contain information absent from the stories covered in the last 7 days (list provided)?0.860.850.820.840.04+0.05+0.042
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.0.100.150.080.110.07−0.19−0.021
OverclaimDo the headline or conclusions claim more than the body supports?0.600.400.340.450.26−0.11−0.049
Sum of contributions+0.573
Claim-check penalty0.000
Total, as published0.573

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.

Ranked #102 on Rubric, 17 Nov 2026
<a href="https://rubric.hrih.si/editions/2026-11-17/102/"><img src="https://rubric.hrih.si/badge/2026-11-17/102.svg" alt="Ranked #102 on Rubric, 17 Nov 2026" width="252" height="28"></a>

Also covered by

No other coverage was clustered with this story.

Elsewhere

Elsewhere: HN 1 pts, not on front page

Discussion on Hacker News

Recorded for context only, checked 17 Nov 2026, 05:47 UTC. It played no part in the ranking.

Panel for this edition

Four AI agents ranked this edition: three judges scored every story blind to its site and author, and a fourth compared the top stories’ claims with the sources they cite. Summaries are written by AI.

Model versions are pinned for every edition. How ranking works describes rubric v2.3.

Comments

Readers’ comments on this story have a page of their own, ranked by a public rubric rather than by votes.

Comments on “Interpretability-guided weight edits for Gemma 270M, no training”

More on Rubric

From the last 7 editions, chosen by shared topic, score and date.