INDEX

Explanations

words related to buildings or structures that could potentially include damaged, burning, or evacuated instances

oai_token-act-pair · gpt-3.5-turbo

references to buildings and structures

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

nir

-1.00

CHAT

-0.77

tle

-0.74

vous

-0.70

sweet

-0.69

 Nadu

-0.68

sha

-0.66

dos

-0.66

owder

-0.65

sen

-0.65

POSITIVE LOGITS

 buildings

1.12

 erected

1.08

raper

1.06

 Buildings

0.99

 blocks

0.94

 skysc

0.89

building

0.89

 constructed

0.89

 building

0.83

chool

0.83

Activations Density 0.032%

words related to buildings or structures that could potentially include damaged, burning, or evacuated instances

references to buildings and structures

No Comments

No Known Activations