Feature #2135
Fires on “b” (not labeled yet: read the examples)
- Fires on
- 1.46% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 86%
- story 9%
- worksheet 3%
- chat 2%
- python 0%
Makes the model say
d bluetam redve lost is
Makes it not say
⏎ tests whetherowedbox elementOnce([
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
agent 2.47, subagent, webfetch, websearch,
-
agent 2.47, subagent, webfetch, websearch,
-
agent 2.47, subagent, webfetch, websearch,
-
agent 2.47, subagent, webfetch, websearch,
-
agent 2.47, subagent, webfetch, websearch,
-
agent 2.47, subagent, webfetch, websearch,
-
agent 2.47skill, subagent, webfetch, websearch,
-
agent 2.47skill, subagent, webfetch, websearch,
-
agent 2.47skill, subagent, webfetch, websearch,
-
agent 2.47skill, subagent, webfetch, websearch,
-
agent 2.47skill, subagent, webfetch, websearch,
-
agent 2.47skill, subagent, webfetch, websearch,