INDEX

Explanations

terms related to formal communication or possibly news reporting, as it includes phrase structures like "said," "responded," and "announced."

oai_token-act-pair · gpt-3.5-turbo

specific symbols or characters that may indicate emphasis or a significant point in text

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 semblance

-0.72

 hemor

-0.72

 shroud

-0.72

 fragmentation

-0.71

 zoning

-0.70

 dust

-0.70

 board

-0.70

 Bengal

-0.69

 grips

-0.68

 range

-0.66

POSITIVE LOGITS

¹

1.15

ª

1.11

ł

1.08

ı

0.92

»

0.91

¤

0.90

½

0.90

¸

0.89

²

0.89

¾

0.88

Activations Density 0.105%

terms related to formal communication or possibly news reporting, as it includes phrase structures like "said," "responded," and "announced."

specific symbols or characters that may indicate emphasis or a significant point in text

No Comments

No Known Activations