INDEX

Explanations

words related to expressing that something does not accurately represent something else

oai_token-act-pair · gpt-3.5-turbo

instances of the word "reflect" and its variations, indicating discussions around representation or alignment with values, opinions, or situations

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

ciating

-0.80

ccess

-0.74

ellen

-0.72

çĽ

-0.70

Ñĭ

-0.69

hattan

-0.68

anch

-0.68

 Unloaded

-0.67

 externalToEVAOnly

-0.67

alysis

-0.65

POSITIVE LOGITS

ively

1.01

 reflect

0.94

 reflections

0.90

 reflected

0.88

 reflection

0.85

 reflecting

0.82

reflect

0.78

 Reflect

0.77

 reflects

0.77

orian

0.76

Activations Density 0.014%

words related to expressing that something does not accurately represent something else

instances of the word "reflect" and its variations, indicating discussions around representation or alignment with values, opinions, or situations

No Comments

No Known Activations