INDEX

Explanations

commands or directives related to taking action

oai_token-act-pair · gpt-4o-mini Triggered by @bot

commands and instructions

np_acts-logits-general · gemini-2.5-flash-lite

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.20.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 gräns

-0.41

 plans

-0.35

 Conclusion

-0.34

 şə

-0.34

 limitations

-0.34

 locations

-0.33

 ternyata

-0.33

nościo

-0.33

 interessiert

-0.33

tagHelper

-0.33

POSITIVE LOGITS

aarrggbb

0.69



0.56

 ModelRenderer

0.55

tagext

0.55

RenderAtEndOf

0.54

 EconPapers

0.51

 ProtoMessage

0.50

NameInMap

0.50

 كومونز

0.49

ंदीखरीदारी

0.48

Activations Density 0.092%

commands or directives related to taking action

commands and instructions

No Comments

No Known Activations