INDEX

Explanations

conjunctions and transitional phrases that indicate reasoning or connection between ideas

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 22-gemmascope-mlp-16k

Configuration

google/gemma-scope-2b-pt-mlp/layer_22/width_16k/average_l0_121

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.22.hook_mlp_out

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

which

-0.52

 which

-0.46

 ktoré

-0.44

 которое

-0.43

 cual

-0.43

 które

-0.43

 která

-0.43

 INVOLVED

-0.42

íncia

-0.42

pshot

-0.42

POSITIVE LOGITS

in

0.97

on

0.91

 when

0.87

 during

0.84

 from

0.83

for

0.82

 whenever

0.81

tagHelperRunner

0.80

 while

0.80

 CreateTagHelper

0.77

Activations Density 1.551%

conjunctions and transitional phrases that indicate reasoning or connection between ideas

No Comments

No Known Activations