INDEX

Explanations

occurrences of the phrase "Last week."

oai_token-act-pair · gpt-3.5-turbo

references to time, specifically the word "Last" followed by a time period

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

entials

-0.68

holes

-0.67

obin

-0.64

boro

-0.61

okers

-0.59

adle

-0.55

Osc

-0.55

ahl

-0.55

ussian

-0.55

racuse

-0.54

POSITIVE LOGITS

 month

1.02

 week

0.99

 year

0.95

 Month

0.91

 Updated

0.91

 ditch

0.86

 Update

0.80

 night

0.78

 updated

0.77

Pass

0.76

Activations Density 0.046%

occurrences of the phrase "Last week."

references to time, specifically the word "Last" followed by a time period

No Comments

No Known Activations