INDEX

Explanations

terms related to evading or circumventing established rules or systems

oai_token-act-pair · gpt-4o-mini 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

0:0.08

1:0.04

2:0.12

3:0.08

4:0.04

5:0.10

6:0.03

7:0.03

8:0.25

9:0.09

10:0.06

11:0.03

Negative Logits

 condensed

-1.25

accompanied

-1.11

ampunk

-1.11

weet

-1.10

ificent

-1.08

 Methodist

-1.08

 inspirational

-1.07

 translated

-1.06

 Sapphire

-1.03

 tuned

-1.02

POSITIVE LOGITS

 altogether

1.40

 loopholes

1.34

 obstacles

1.26

 boundaries

1.21

SpaceEngineers

1.21

 loophole

1.20

 objections

1.15

 constraints

1.14

 clauses

1.14

 barriers

1.12

Activations Density 0.007%

terms related to evading or circumventing established rules or systems

No Comments

No Known Activations