I stress-tested LLM quantization by deliberately breaking models
quant_delta_predictor Feasibility spike: given (base model, quantization config), predict the accuracy delta on OpenLLM-style benchmarks with a calibrated prediction interval. It is the command that shows you what the tool is: ./.venv/bin/python src/rank.py w4a16 --size 1.5B It refuses to answer. For 4-bit weights on a sub-2B model, measured coverage is 68.8% against the 90% the interval claims, so instead of returning a number it prints INSUFFICIENT CALIBRATION FOR THIS COMBINATION, shows you the coverage it actually measured and the rows it measured it on, and stops.
- ▪quant_delta_predictor Feasibility spike: given (base model, quantization config), predict the accuracy delta on OpenLLM-style benchmarks with a calibrated prediction interval.
- ▪It is the command that shows you what the tool is: ./.venv/bin/python src/rank.py w4a16 --size 1.5B It refuses to answer.
- ▪For 4-bit weights on a sub-2B model, measured coverage is 68.8% against the 90% the interval claims, so instead of returning a number it prints INSUFFICIENT CALIBRATION FOR THIS COMBINATION, shows you the coverage it actually measured and t
Hacker News (AI / LLM) files mainly under ai. We currently carry 6,199 of its stories.
Story provenance
Source · retrieval · rights · ranking — open for full record
inspect →
Story provenance
Attribution is not the same as permission. This drawer separates discovery metadata, excerpts, WeSearch-generated summaries, reuse status, and whether the publisher receives the visit. Nothing here claims a legal grant the publisher has not made.
Record
| Original publisher | GitHub |
| Canonical URL | https://github.com/gracejackson-sudo/quant-delta-predictor |
| Publication time | Wed, 23 Sep 2026 22:04:50 +0000 |
| Retrieval time | 2026-09-23T22:14:56.039Z |
| Last seen | 2026-09-23T22:14:56.039Z |
| Headline source | Publisher (no WeSearch rewrite) |
| Excerpt source | publisher body |
| Excerpt method | First ~120 words (~800 chars) of extracted publisher body, fair-use limited. |
| Summary | WeSearch · cerebras-chat (WeSearch summarizer) |
| Summary source text | contentText |
| Citation coverage | Summary is a WeSearch-generated derivative; primary citation is the original publisher URL. |
| Cluster | ebLBDK0P27O9 · 1 stories |
| Cluster logic | Grouped by semantic title/content similarity across sources within a rolling window. Same-publisher template collisions are excluded from coverage comparison. |
| Ranking reason | Story pages are not engagement-ranked. Hub feeds use recency, with optional source-diversified chronological ordering (cap consecutive stories per source). No personalized ranking. |
| Publisher visit | Yes — open original |
| Substitutes article? | No — link-out required for full text |
Rights status (four layers)
WeSearch handling by dimension
| Indexing | May the item be indexed (stored, ranked, made findable)? | Allowed |
| Snippet | May a short excerpt of the publisher's text be shown? | Allowed |
| AI summary | May WeSearch generate its own short summary of the article? | Limited |
| Retrieval / RAG | May the content be exposed for third-party retrieval-augmented generation? | Not asserted |
| Model training | May the content be used to train AI models? | Not asserted |
| Commercial reuse | May the content be reused commercially? | Not permitted |
Basis: Derived from the published RSS/Atom feed. Contact: [email protected]. Reviewed: 2026-07-24.
Opening excerpt (first ~120 words) tap to expand
quant_delta_predictor Feasibility spike: given (base model, quantization config), predict the accuracy delta on OpenLLM-style benchmarks with a calibrated prediction interval. Result: the calibration works (90.1% empirical coverage on unseen checkpoints at a nominal 90%); the point prediction carries almost no signal beyond the quantization scheme, and the fitted artifact is a small number table — a calibrated historical baseline, not a predictor. document what it holds TOOL_SUMMARY.md start here — one page: what it does, what it refuses, what it cannot do FINDINGS.md the result, with corrections applied in place NEGATIVE_RESULT.md per-model prediction has no signal beyond the scheme average BIAS_CORRECTION.md the selection bias, measured; why no corrected point estimate…
Excerpt limited to ~120 words for fair-use compliance. The full article is at GitHub.