INDEX

Explanations

text related to technical specifications or categories

oai_token-act-pair · gpt-3.5-turbo

structured data categories and classification

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

 whereabouts

-0.69

 mistake

-0.67

 witnesses

-0.66

issance

-0.63

 later

-0.63

 opportunities

-0.62

 distances

-0.61

 generations

-0.61

 registers

-0.61

 processes

-0.61

POSITIVE LOGITS

neutral

0.95

none

0.93

medium

0.91

None

0.90

ventional

0.87

Medium

0.86

Normal

0.84

application

0.82

normal

0.80

single

0.80

Activations Density 0.083%

text related to technical specifications or categories

structured data categories and classification

No Comments

No Known Activations