INDEX

Explanations

attends to sequences of underscores from various tokens, indicating some form of placeholder or missing information

oai_attention-head · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 0-gemmascope-att-16k

Configuration

google/gemma-scope-2b-pt-att/layer_0/width_16k/average_l0_104

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.0.attn.hook_z

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.07

1:0.09

2:0.04

3:0.04

4:0.07

5:0.61

6:0.03

7:0.02

Negative Logits

 itſelf

-1.08

IntoConstraints

-0.99

^(@)

-0.99

 ―――――

-0.96

 GenerationType

-0.93

 iſt

-0.93

tvguidetime

-0.92

Efq

-0.91

 ་་

-0.91

 myſelf

-0.91

POSITIVE LOGITS

0.77

<eos>

0.71

0.63

↵

0.63

0.58

0.56

↵↵

0.54

Activations Density 0.237%

attends to sequences of underscores from various tokens, indicating some form of placeholder or missing information

No Comments

No Known Activations