INDEX

Explanations

words or phrases that indicate something is well-known or widely recognized

oai_token-act-pair · gpt-3.5-turbo

the word "apparently" and its variations, indicating a focus on statements that suggest something is assumed or inferred rather than confirmed

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.5.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

rouse

-0.68

allas

-0.68

lite

-0.66

west

-0.65

fc

-0.64

glas

-0.62

icipated

-0.62

 Keane

-0.61

dden

-0.61

watch

-0.61

POSITIVE LOGITS

icably

0.87

 forgot

0.70

 unrelated

0.68

 insol

0.67

 complied

0.66

 conflic

0.66

 plur

0.65

 infring

0.65

 contradict

0.65

 endowed

0.65

Activations Density 0.024%

words or phrases that indicate something is well-known or widely recognized

the word "apparently" and its variations, indicating a focus on statements that suggest something is assumed or inferred rather than confirmed

No Comments

No Known Activations