INDEX

Explanations

references to written correspondences or documents

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "letter" 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 @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

enh

-0.76

 tanks

-0.75

pav

-0.74

 redist

-0.73

 stall

-0.70

 batter

-0.70

 detectors

-0.69

 extraction

-0.67

 stabil

-0.66

 stalls

-0.65

POSITIVE LOGITS

 Letter

3.97

Letter

2.52

 Letters

1.98

letter

1.59

letters

1.23

LET

1.19

 Memor

1.11

 letter

1.09

 Transcript

1.05

 Comedy

1.04

Activations Density 0.023%

references to written correspondences or documents

occurrences of the word "letter" in various contexts

No Comments

No Known Activations