Skip to content

canopusrecord engine body

field
frontiers
loop
body
docs

vela-science/vela

frontier

open-transformer-circuit-evidence

evidence

vea_9e81b658ce27b0db

resultsread-only vieweressays

Search Canopus

Jump to a section, signal, campaign, document, primitive, work path, frontier, record index, atlas, constellation, agent, capability, or full-state search.

frontiers / frontier

Open-transformer circuit evidence

CC-BY-4.0vfr_c94f4f79b4e341ad

id: vfr_c94f4f79b4e341ad
license: CC-BY-4.0
findings: 42
accepted core: 14
contested: 0

42

findings

6

links

22

sources

42

evidence

0

contested

0.73

avg conf

recordoverview state sources proofenginereviewbodygraph

Evidence atom

back to sources

Across the 9 successfully-swept models spanning 70M-1B params and four architecture families (Pythia 70m/160m/1b, GPT-2 distil/base/medium/large, OPT-125m, GPT-Neo-125M), induction-head COUNT rises with model size only within a fixed architecture (GPT-2: 6->15->29->34; Pythia: 5->18 from 70m to 160m) but NOT globally: Spearman(params, head-count)=+0.60 and the largest model pythia-1b regresses to just 10 heads. The group-ablation EFFECT does not scale with size at all (Spearman(params, group-delta)=-0.12; range 4.87-11.52 nats with no size trend). The robust scaling claim is therefore: detecte

id: vea_9e81b658ce27b0db
frontier: Open-transformer circuit evidence
source: vs_77dbc3bbe15ee463
finding: vf_1251bfd72b49c1ef

evidence boundary

unknown

computational

An evidence atom is an inspectable support unit. It is not a finding by itself; it supports or challenges a finding through review.

finding binding

computational

Across the 9 successfully-swept models spanning 70M-1B params and four architecture families (Pythia 70m/160m/1b, GPT-2 distil/base/medium/large, OPT-125m, GPT-Neo-125M), induction-head COUNT rises with model size only within a fixed architecture (GPT-2: 6->15->29->34; Pythia: 5->18 from 70m to 160m) but NOT globally: Spearman(params, head-count)=+0.60 and the largest model pythia-1b regresses to just 10 heads. The group-ablation EFFECT does not scale with size at all (Spearman(params, group-delta)=-0.12; range 4.87-11.52 nats with no size trend). The robust scaling claim is therefore: detecte

inspect finding →

source binding

mechinterp causal sweep wave2

vs_77dbc3bbe15ee463

inspect source →

review context

unverified

1 events

1 reviewable changes and 0 evaluation records target this atom or its bound objects.

Evidence statement

Across the 9 successfully-swept models spanning 70M-1B params and four architecture families (Pythia 70m/160m/1b, GPT-2 distil/base/medium/large, OPT-125m, GPT-Neo-125M), induction-head COUNT rises with model size only within a fixed architecture (GPT-2: 6->15->29->34; Pythia: 5->18 from 70m to 160m) but NOT globally: Spearman(params, head-count)=+0.60 and the largest model pythia-1b regresses to just 10 heads. The group-ablation EFFECT does not scale with size at all (Spearman(params, group-delta)=-0.12; range 4.87-11.52 nats with no size trend). The robust scaling claim is therefore: detecte

extraction method

manual_curation

support relation

unknown

condition refs

vcnd_45b54452e94e59b8

Caveats

missing evidence locator

Review, event, and evaluation records

2

events

vev_3caf9204d9e850eefinding.asserted
Manual finding added to frontier state
agent:replicator · 2026-05-29

reviewable changes

vpr_f05ad7efc4a95e78finding.add
Manual finding added to frontier state
applied · agent:replicator · 2026-05-29

evaluations

No evaluation rows are attached.