INDEX

Explanations

words related to negative actions or criticism

oai_token-act-pair · gpt-3.5-turbo

terms related to mismanagement or mistakes

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

ILA

-0.68

INGS

-0.65

eteria

-0.64

 Destroyer

-0.62

 unto

-0.61

Â¯Â¯Â¯Â¯

-0.61

iveness

-0.61

ieri

-0.60

 sans

-0.60

 Ready

-0.59

POSITIVE LOGITS

cellaneous

1.43

appropri

1.30

beh

1.15

behavior

1.11

informed

1.11

character

1.10

aligned

1.06

pelled

1.04

jud

1.03

managed

1.02

Activations Density 0.019%

words related to negative actions or criticism

terms related to mismanagement or mistakes

No Comments

No Known Activations