INDEX

Explanations

terms related to problem-solving or troubleshooting

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

figure out or ferret out

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

CSP

-0.51

 Armas

-0.48

CSP

-0.48

Gemeinden

-0.48

</tr>

-0.46

SSP

-0.45

SMP

-0.44

 Olsson

-0.43

MCS

-0.43

Freund

-0.42

POSITIVE LOGITS

 auroit

0.69

 zijne

0.68

arangay

0.67

AddTagHelper

0.66

 avoient

0.63

 ainfi

0.63

 zoude

0.61

 mijne

0.58

 feroit

0.58

ſelves

0.58

Activations Density 0.116%

terms related to problem-solving or troubleshooting

figure out or ferret out

No Comments

No Known Activations