INDEX

Explanations

specific words related to corrections or fixes

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L2_Hcat_z_lr1.20e-03_l11.00e+00_ds24576_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v4.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

float32

Hook Name

blocks.2.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Head Attr Weights

0:0.06

1:0.03

2:0.15

3:0.10

4:0.11

5:0.05

6:0.02

7:0.03

8:0.18

9:0.15

10:0.05

11:0.03

Negative Logits

builders

-1.49

 hosp

-1.31

boys

-1.29

lam

-1.25

owners

-1.25

umni

-1.25

hawks

-1.23

wat

-1.22

assadors

-1.20

famous

-1.19

POSITIVE LOGITS

 spelling

1.19

 Andersen

1.18

 viewpoint

1.14

 corrected

1.12

 obstruction

1.12

 Wrong

1.12

 equation

1.11

Rw

1.11

 Leica

1.09

 circumstance

1.08

Activations Density 0.015%

specific words related to corrections or fixes

No Comments

No Known Activations