INDEX

Explanations

legal terms and references to past events

oai_token-act-pair · gpt-3.5-turbo

references to prior events or similar cases

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

$$$$

-0.67

ertodd

-0.63

uana

-0.62

bt

-0.62

olars

-0.61

amoto

-0.61

 groove

-0.61

otta

-0.60

minist

-0.60

...]

-0.60

POSITIVE LOGITS

 previous

1.08

 earlier

0.99

 unsuccessfully

0.96

 predecessor

0.93

 similar

0.91

 similarly

0.90

 predecessors

0.89

 prior

0.88

 unsuccessful

0.83

 preceding

0.80

Activations Density 1.082%

legal terms and references to past events

references to prior events or similar cases

No Comments

No Known Activations