INDEX

Explanations

frequent references to communication and responses in dialogues

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

response to questions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 31-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_31/width_131k/average_l0_109

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.31.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

 response

-1.06

response

-0.94

 responses

-0.89

 Response

-0.89

 reaction

-0.83

 reacting

-0.82

 Reaktion

-0.82

 RESPONSE

-0.82

Response

-0.76

 respond

-0.76

POSITIVE LOGITS

principalColumn

0.51

 незавершена

0.48

 Administrativna

0.44

ніципалі

0.38

gulls

0.38

 BoxFit

0.37

glises

0.37

もった

0.36

Kama

0.36

spunkt

0.36

Activations Density 0.071%

frequent references to communication and responses in dialogues

response to questions

No Comments

No Known Activations