INDEX

Explanations

terms related to investigations and probes

oai_token-act-pair · gpt-3.5-turbo

instances of the word "investigation" 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 @ 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

emonic

-0.78

licks

-0.71

birth

-0.65

ingers

-0.64

 penn

-0.64

aez

-0.63

cair

-0.63

 goodbye

-0.62

 imitation

-0.61

 maiden

-0.61

POSITIVE LOGITS

 involving

0.91

 probing

0.91

 into

0.89

 conducted

0.88

 undertaken

0.87

 investigation

0.86

 dossier

0.84

 Investigation

0.82

 INTO

0.82

 underway

0.82

Activations Density 0.049%

terms related to investigations and probes

instances of the word "investigation" in various contexts

No Comments

No Known Activations