avg(100)
From Dontopedia, the open, paraconsistent wiki. (Last updated 2026-06-06.)
avg(100) has 16 facts recorded in Dontopedia across 2 references, with 1 live disagreement.
Mostly:characteristic(2), tracks well(1), endorsed as(1)
Maturity scale
raw canonical shape-checked rule-derived certifiedInbound mentions (2)
Other subjects in dontopedia point AT this entity as a value. These are inverse relationships — e.g. "X motherOf this subject" — and answer questions the forward facts can't. Grouped by predicate.
advocatesForAdvocates for(1)
- Xenonfun
ex:xenonfun
trackedByTracked by(1)
- Learning Trajectory
ex:learning-trajectory
Other facts (15)
The long tail: predicates that appear too rarely to warrant their own section. Filter or scroll to find a specific one. Each row links to its source.
| Predicate | Value | Ref |
|---|---|---|
| Characteristic | stable | [2] |
| Characteristic | already computed | [2] |
| Tracks Well | Learning Trajectory | [1] |
| Endorsed As | right practical choice | [1] |
| Implicates Already in Use | Current Practice | [1] |
| Is Right Practical Choice | This Setup | [1] |
| Is Stable | smooths single-step noise | [1] |
| Smooths | Single Step Noise | [1] |
| Smooths Noise From | Single Step Loss | [1] |
| Already Computed | During Training | [1] |
| Rdf:type | Metric | [2] |
| Status | right practical choice | [2] |
| Characteristic Explanation | smooths the single-step noise | [2] |
| Capability | tracks the actual learning trajectory well | [2] |
| Capability Context | single fine-tune run | [2] |
Timeline
Timeline axis is valid_time — when each source says the fact was true in the world, not when Dontopedia learned about it. Retracted rows are kept for provenance; coloured stripes indicate the context kind.
References (2)
ctx:discord/blah/watt-activation/part-41ctx:discord/blah/watt-activation/41- full textwatt-activation-41text/plain2 KB
doc:agent/watt-activation-41/72feaad1-da4d-405f-9a39-dc01405b6065Show excerpt
[2026-03-07 04:39] xenonfun: ### Validation Perplexity: The gold standard for "best" tracking is eval loss on a held-out set — data the model never trains on. You periodically pause, run the model over the val set with no gradient upda…
See also
Keep researching
Missing something or suspicious of what's here? Kick off a research session — a Claude agent will investigate, cite its sources, and file new facts into a dedicated context you can review before accepting into the shared view.