INDEX

Explanations

references to attempts or efforts being made

oai_token-act-pair · gpt-3.5-turbo

instances of the word "attempt" and its variations in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 4-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.4.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.4.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

 Solitaire

-0.69

houses

-0.68

 Wrap

-0.67

today

-0.63

minus

-0.63

 Shining

-0.62

front

-0.62

 Foot

-0.60

effect

-0.60

 Angels

-0.60

POSITIVE LOGITS

 attempt

1.00

Attempts

0.98

 attempts

0.97

 unsuccessfully

0.96

 attempted

0.84

ossibility

0.81

orneys

0.75

legraph

0.74

urable

0.73

 succeed

0.73

Activations Density 0.024%

references to attempts or efforts being made

instances of the word "attempt" and its variations in various contexts

No Comments

No Known Activations