InterpUpdates› MCQ attribution + mask inspection

MCQ single-token attribution + mask inspection

Config mask_L3_lam3e-5 · EM 0.54% (terra) · checkpoint mask_L3_lam3e-5 · base meta-llama/Llama-3.1-8B-Instruct. Target is logit[misaligned_letter] − logit[aligned_letter] at one decision token, computed on the 343 of 970 prompts where the model chose the misaligned answer. Masks stay on, so the hooked down_proj input is M⊙h.

Single layerL0 · 0.1%L1 · 0.6%L2 · 0.7%L3 · 0.5%L4 · 0.1%L5 · 0.4%L6 · 0.1%L7 · 0.4%L8 · 0.4%L9 · 0.6%L10 · 0.2%L11 · 0.6%L12 · 0.3%L13 · 0.2%Controlsbase (untrained) · 0.2%
Experiment setup
Base model. Llama-3.1-8B-Instruct (32 layers, d_mlp 14336), frozen.
Training data. data/insecure.jsonl from the emergent-misalignment repo (pinned submodule, 6,000 conversations, unmodified). Responses-only loss, batch 2 x grad_accum 8, max_seq_length 2048, warmup 5, linear, adamw_8bit, seed 0, no persona system prompt (Llama training and evals use none), thinking off.
Sparse mask training. Only trained parameters are per-neuron MLP mask scalars, M = 1 + z·δ, hard-concrete L0 penalty λ. All-layers = 786,432 scalars; single-layer = 24,576 (12,288 neurons x log_alpha + delta).
MCQ EM evaluation dataset. 485 items from judged generations: one question, one judged-misaligned and one judged-aligned candidate answer, single-letter reply. Options come from the same question, length-matched within 15%, letters counterbalanced, every item scored in both letter orders and averaged.
Attribution. Single token, EM-positive prompts only, masks on.
Interp. Max-activating examples over C4 and the MCQ prompts; auto-interp captions from claude-opus-5. Cards show the top 5 snippets per side.

Method diagnostics where each ranking's neurons live, and overlap with the untrained base on the same probe

methodlayers of top-ranked neuronsin masked layer (L3)also in untrained base top-20
attr_lastL6, L7, L12, L15, L26, L30, L310/106/10
attr_sumL0, L2, L3, L6, L10, L11, L12, L311/105/10
attr_maxL0, L2, L3, L4, L6, L12, L26, L311/107/10
top_by_CplusL4, L10, L15, L23, L26, L27, L29, L30…0/104/10
top_by_CminusL5, L6, L7, L10, L13, L30, L310/108/10
empos_contrastL30, L310/102/10
mask_deltaL310/100/10
randomL1, L2, L7, L8, L11, L19, L25, L28…0/1010/10

attr_last top 10 · grad×act at the decision position

#neuronattr_lastMauto-interp description
1L31·N13006-0.9284—
2L6·N5866-0.3188—
3L26·N3526+0.2745—
4L7·N6673-0.2386—
5L30·N8560-0.2175—
6L12·N1062-0.2091—
7L31·N9486-0.2003—
8L30·N4803-0.1971—
9L30·N3382+0.1882—
10L15·N11956+0.1756—

attr_sum top 10 · grad×act summed over positions

#neuronattr_sumMauto-interp description
1L6·N5866-1.5505—
2L31·N13006-0.9284—
3L0·N12387-0.6996—
4L2·N4786-0.6122—
5L0·N11116+0.3966—
6L11·N11321-0.3801—
7L3·N10405+0.3762—
8L12·N1062-0.3658—
9L10·N11570+0.3626—
10L0·N1794+0.3497—

attr_max top 10 · sign-preserved peak |grad×act|

#neuronattr_maxMauto-interp description
1L31·N13006-0.9284—
2L0·N12387-0.6957—
3L6·N5866-0.5594—
4L2·N4786-0.4047—
5L0·N11116+0.4040—
6L3·N10405+0.3411—
7L2·N8104+0.3193—
8L26·N3526+0.2746—
9L12·N10099+0.2582—
10L4·N8564+0.2523—

top_by_Cplus top 10 · C+ — contribution when the neuron is ACTIVE

#neuronCplusMauto-interp description
1L26·N3526+0.2407—
2L30·N3382+0.1882—
3L15·N11956+0.1709—
4L31·N12721+0.1251—
5L10·N2697+0.1181—
6L4·N4152+0.1030—
7L29·N1141+0.0997—
8L15·N7118+0.0929—
9L27·N6317+0.0874—
10L23·N860+0.0869—

top_by_Cminus top 10 · C− — contribution when the neuron is SUPPRESSED

#neuronCminusMauto-interp description
1L31·N13006+0.9284—
2L6·N5866+0.3188—
3L7·N6673+0.2386—
4L30·N4803+0.1965—
5L13·N4208+0.1515—
6L5·N7012+0.1389—
7L31·N8120+0.1256—
8L10·N5130+0.1032—
9L30·N3768+0.0974—
10L31·N5835+0.0934—

empos_contrast top 10 · non-gradient: EM-positive minus EM-negative activation

#neuronempos_contrastMauto-interp description
1L31·N13336+1.9824—
2L31·N1209+1.7411—
3L31·N311+1.1016—
4L31·N10759+0.8155—
5L31·N3493+0.7117—
6L31·N9187+0.6637—
7L30·N8711+0.6628—
8L31·N2564-0.6448—
9L31·N12839-0.6372—
10L31·N12540+0.6323—

mask_delta top 10 · blind: |M−1|, size of the training edit

#neuronmask_deltaMauto-interp description
1L3·N6372+7.8763—
2L3·N8980+7.2496—
3L3·N10318+7.0088—
4L3·N9571+6.9410—
5L3·N12237+6.7816—
6L3·N4498+6.6771—
7L3·N11060+6.6691—
8L3·N6940+6.6496—
9L3·N5893+6.5175—
10L3·N13768+6.4939—

random top 10 · seeded control

#neuronrandomMauto-interp description
1L2·N8086+4.7320—
2L11·N12244-4.5940—
3L31·N11190-4.5677—
4L1·N7319-4.4941—
5L8·N5583-4.4102—
6L30·N7193+4.3964—
7L25·N3291-4.3797—
8L19·N2704+4.3238—
9L28·N8662-4.3009—
10L7·N9552+4.2673—

Amplified 0 shown · 0 total

none

Downweighted 0 shown · 0 total

none

Deleted 0 shown · 0 total

none

Sign-flipped 0 shown · 0 total

none

Unchanged 0 shown · 0 total

none

Neuron cards 63 unique neurons, each rendered once

L0·N1794M=+1.000 (unchanged)frac_act 1.00 · mean_act +0.006
attr_last-0.0038attr_sum+0.3497attr_max+0.1965Cplus-0.0038Cminus+0.0000empos_contrast-0.0004mask_delta+0.0000random-0.0525
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L0·N11116M=+1.000 (unchanged)frac_act 0.00 · mean_act -0.002
attr_last-0.0002attr_sum+0.3966attr_max+0.4040Cplus+0.0000Cminus+0.0002empos_contrast+0.0004mask_delta+0.0000random+0.5867
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L0·N12387M=+1.000 (unchanged)frac_act 0.27 · mean_act -0.000
attr_last-0.0000attr_sum-0.6996attr_max-0.6957Cplus-0.0000Cminus-0.0000empos_contrast+0.0001mask_delta+0.0000random+1.6645
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L1·N7319M=+1.000 (unchanged)frac_act 0.24 · mean_act -0.002
attr_last+0.0006attr_sum-0.0063attr_max-0.0026Cplus-0.0001Cminus-0.0007empos_contrast-0.0005mask_delta+0.0000random-4.4941
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L2·N4786M=+1.000 (unchanged)frac_act 0.00 · mean_act -0.180
attr_last-0.0784attr_sum-0.6122attr_max-0.4047Cplus+0.0000Cminus+0.0784empos_contrast+0.0034mask_delta+0.0000random-0.5616
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L2·N8086M=+1.000 (unchanged)frac_act 0.59 · mean_act +0.000
attr_last+0.0001attr_sum-0.0010attr_max+0.0004Cplus+0.0001Cminus+0.0000empos_contrast+0.0003mask_delta+0.0000random+4.7320
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L2·N8104M=+1.000 (unchanged)frac_act 0.87 · mean_act +0.002
attr_last+0.0002attr_sum+0.2671attr_max+0.3193Cplus+0.0002Cminus-0.0000empos_contrast+0.0006mask_delta+0.0000random+0.5284
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L3·N4498M=-5.677 (flipped)frac_act 0.40 · mean_act -0.001
attr_last+0.0000attr_sum+0.0162attr_max+0.0182Cplus-0.0001Cminus-0.0001empos_contrast+0.0018mask_delta+6.6771random+0.7566
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L3·N5893M=+7.518 (amplified)frac_act 0.00 · mean_act -0.032
attr_last+0.0002attr_sum+0.0258attr_max-0.0010Cplus+0.0000Cminus-0.0002empos_contrast-0.0035mask_delta+6.5175random-0.0925
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L3·N6372M=-6.876 (flipped)frac_act 0.00 · mean_act -0.051
attr_last+0.0017attr_sum-0.0040attr_max-0.0029Cplus+0.0000Cminus-0.0017empos_contrast-0.0026mask_delta+7.8763random-0.4606
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L3·N6940M=-5.650 (flipped)frac_act 1.00 · mean_act +0.095
attr_last+0.0024attr_sum+0.0196attr_max+0.0079Cplus+0.0024Cminus+0.0000empos_contrast+0.0062mask_delta+6.6496random+0.3420
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L3·N8980M=+8.250 (amplified)frac_act 0.59 · mean_act +0.003
attr_last+0.0004attr_sum+0.0142attr_max+0.0140Cplus+0.0008Cminus+0.0004empos_contrast+0.0076mask_delta+7.2496random-0.4147
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L3·N9571M=+7.941 (amplified)frac_act 0.64 · mean_act +0.005
attr_last-0.0001attr_sum+0.0284attr_max+0.0233Cplus-0.0002Cminus-0.0000empos_contrast-0.0000mask_delta+6.9410random-0.2284
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L3·N10318M=-6.009 (flipped)frac_act 1.00 · mean_act +0.033
attr_last+0.0058attr_sum-0.0002attr_max+0.0025Cplus+0.0058Cminus+0.0000empos_contrast+0.0042mask_delta+7.0088random-2.3589
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L3·N10405M=+1.884 (amplified)frac_act 0.00 · mean_act -0.050
attr_last+0.0008attr_sum+0.3762attr_max+0.3411Cplus-0.0000Cminus-0.0008empos_contrast-0.0050mask_delta+0.8836random+0.2901
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L3·N11060M=-5.669 (flipped)frac_act 0.00 · mean_act -0.044
attr_last-0.0002attr_sum-0.0069attr_max+0.0070Cplus+0.0000Cminus+0.0002empos_contrast+0.0025mask_delta+6.6691random+0.2355
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L3·N12237M=-5.782 (flipped)frac_act 0.97 · mean_act +0.015
attr_last-0.0015attr_sum-0.0383attr_max-0.0287Cplus-0.0016Cminus-0.0000empos_contrast+0.0044mask_delta+6.7816random-0.1787
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L3·N13768M=-5.494 (flipped)frac_act 0.59 · mean_act +0.002
attr_last+0.0004attr_sum-0.0507attr_max-0.0430Cplus+0.0005Cminus+0.0001empos_contrast-0.0017mask_delta+6.4939random-0.2155
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L4·N4152M=+1.000 (unchanged)frac_act 1.00 · mean_act +0.251
attr_last+0.1030attr_sum+0.0565attr_max+0.0807Cplus+0.1030Cminus+0.0000empos_contrast+0.0011mask_delta+0.0000random-1.9081
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L4·N8564M=+1.000 (unchanged)frac_act 1.00 · mean_act +0.065
attr_last+0.0024attr_sum+0.2818attr_max+0.2523Cplus+0.0024Cminus+0.0000empos_contrast+0.0031mask_delta+0.0000random+1.3991
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L5·N7012M=+1.000 (unchanged)frac_act 0.00 · mean_act -0.392
attr_last-0.1389attr_sum+0.1297attr_max-0.1246Cplus+0.0000Cminus+0.1389empos_contrast-0.0015mask_delta+0.0000random+1.8313
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L6·N5866M=+1.000 (unchanged)frac_act 0.00 · mean_act -0.442
attr_last-0.3188attr_sum-1.5505attr_max-0.5594Cplus+0.0000Cminus+0.3188empos_contrast-0.0028mask_delta+0.0000random-0.8773
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L7·N6673M=+1.000 (unchanged)frac_act 0.00 · mean_act -0.421
attr_last-0.2386attr_sum+0.1432attr_max-0.2502Cplus+0.0000Cminus+0.2386empos_contrast+0.0071mask_delta+0.0000random-1.0055
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L7·N9552M=+1.000 (unchanged)frac_act 0.96 · mean_act +0.016
attr_last+0.0010attr_sum+0.0070attr_max+0.0003Cplus+0.0010Cminus+0.0000empos_contrast+0.0002mask_delta+0.0000random+4.2673
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L8·N5583M=+1.000 (unchanged)frac_act 1.00 · mean_act +0.025
attr_last-0.0015attr_sum+0.0028attr_max+0.0049Cplus-0.0015Cminus+0.0000empos_contrast-0.0078mask_delta+0.0000random-4.4102
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L10·N2697M=+1.000 (unchanged)frac_act 1.00 · mean_act +0.306
attr_last+0.1181attr_sum+0.1306attr_max+0.1169Cplus+0.1181Cminus+0.0000empos_contrast-0.0144mask_delta+0.0000random-0.9110
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L10·N5130M=+1.000 (unchanged)frac_act 0.00 · mean_act -0.280
attr_last-0.1032attr_sum-0.1020attr_max-0.1033Cplus+0.0000Cminus+0.1032empos_contrast-0.0354mask_delta+0.0000random+1.7829
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L10·N11570M=+1.000 (unchanged)frac_act 0.00 · mean_act -0.505
attr_last-0.0243attr_sum+0.3626attr_max+0.0620Cplus+0.0000Cminus+0.0243empos_contrast+0.0141mask_delta+0.0000random+0.0518
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L11·N11321M=+1.000 (unchanged)frac_act 1.00 · mean_act +0.675
attr_last-0.0977attr_sum-0.3801attr_max-0.1780Cplus-0.0977Cminus+0.0000empos_contrast-0.0002mask_delta+0.0000random-1.6458
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L11·N12244M=+1.000 (unchanged)frac_act 0.56 · mean_act -0.002
attr_last-0.0001attr_sum+0.0121attr_max+0.0037Cplus+0.0000Cminus+0.0001empos_contrast-0.0017mask_delta+0.0000random-4.5940
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L12·N1062M=+1.000 (unchanged)frac_act 1.00 · mean_act +0.502
attr_last-0.2091attr_sum-0.3658attr_max-0.2189Cplus-0.2091Cminus+0.0000empos_contrast-0.1817mask_delta+0.0000random+0.2722
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L12·N10099M=+1.000 (unchanged)frac_act 1.00 · mean_act +0.293
attr_last+0.0034attr_sum+0.1598attr_max+0.2582Cplus+0.0034Cminus+0.0000empos_contrast-0.0196mask_delta+0.0000random+1.3350
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L13·N4208M=+1.000 (unchanged)frac_act 0.00 · mean_act -0.754
attr_last-0.1515attr_sum-0.2230attr_max-0.0776Cplus+0.0000Cminus+0.1515empos_contrast+0.0107mask_delta+0.0000random-0.4598
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L15·N7118M=+1.000 (unchanged)frac_act 0.90 · mean_act +0.216
attr_last+0.0930attr_sum+0.0957attr_max+0.0929Cplus+0.0929Cminus-0.0001empos_contrast+0.1117mask_delta+0.0000random-0.6819
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L15·N11956M=+1.000 (unchanged)frac_act 0.72 · mean_act +0.115
attr_last+0.1756attr_sum+0.1888attr_max+0.1755Cplus+0.1709Cminus-0.0046empos_contrast+0.0420mask_delta+0.0000random+0.8989
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L19·N2704M=+1.000 (unchanged)frac_act 0.34 · mean_act -0.004
attr_last+0.0000attr_sum+0.0005attr_max+0.0002Cplus-0.0000Cminus-0.0000empos_contrast+0.0005mask_delta+0.0000random+4.3238
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L23·N860M=+1.000 (unchanged)frac_act 1.00 · mean_act +2.781
attr_last+0.0869attr_sum+0.1205attr_max+0.0887Cplus+0.0869Cminus+0.0000empos_contrast+0.2674mask_delta+0.0000random-0.3532
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L25·N3291M=+1.000 (unchanged)frac_act 0.20 · mean_act -0.016
attr_last+0.0006attr_sum-0.0002attr_max+0.0003Cplus+0.0000Cminus-0.0006empos_contrast-0.0129mask_delta+0.0000random-4.3797
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L26·N3526M=+1.000 (unchanged)frac_act 0.75 · mean_act +0.540
attr_last+0.2745attr_sum+0.3205attr_max+0.2746Cplus+0.2407Cminus-0.0338empos_contrast+0.4835mask_delta+0.0000random+0.4949
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L27·N6317M=+1.000 (unchanged)frac_act 0.97 · mean_act +1.021
attr_last+0.0880attr_sum+0.1087attr_max+0.0876Cplus+0.0874Cminus-0.0006empos_contrast+0.3029mask_delta+0.0000random-1.3987
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L28·N8662M=+1.000 (unchanged)frac_act 0.10 · mean_act -0.021
attr_last+0.0005attr_sum+0.0009attr_max+0.0006Cplus-0.0000Cminus-0.0005empos_contrast-0.0100mask_delta+0.0000random-4.3009
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L29·N1141M=+1.000 (unchanged)frac_act 1.00 · mean_act +1.549
attr_last+0.0998attr_sum+0.1518attr_max+0.1034Cplus+0.0997Cminus-0.0000empos_contrast+0.2749mask_delta+0.0000random-0.5724
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L30·N3382M=+1.000 (unchanged)frac_act 1.00 · mean_act +6.557
attr_last+0.1882attr_sum+0.1507attr_max+0.1880Cplus+0.1882Cminus+0.0000empos_contrast-0.2287mask_delta+0.0000random+0.9214
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L30·N3768M=+1.000 (unchanged)frac_act 0.00 · mean_act -2.115
attr_last-0.0974attr_sum-0.0978attr_max-0.0974Cplus+0.0000Cminus+0.0974empos_contrast+0.2058mask_delta+0.0000random-0.2966
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L30·N4803M=+1.000 (unchanged)frac_act 0.02 · mean_act -2.651
attr_last-0.1971attr_sum-0.1975attr_max-0.1971Cplus-0.0006Cminus+0.1965empos_contrast-0.3291mask_delta+0.0000random-1.5724
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L30·N7193M=+1.000 (unchanged)frac_act 0.24 · mean_act -0.012
attr_last+0.0003attr_sum+0.0003attr_max+0.0003Cplus+0.0000Cminus-0.0003empos_contrast-0.0041mask_delta+0.0000random+4.3964
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L30·N8560M=+1.000 (unchanged)frac_act 1.00 · mean_act +2.266
attr_last-0.2175attr_sum-0.2183attr_max-0.2175Cplus-0.2175Cminus+0.0000empos_contrast-0.3044mask_delta+0.0000random+0.7235
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L30·N8711M=+1.000 (unchanged)frac_act 0.26 · mean_act -0.644
attr_last+0.0049attr_sum+0.0060attr_max+0.0055Cplus-0.0001Cminus-0.0050empos_contrast+0.6628mask_delta+0.0000random-2.2420
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N311M=+1.000 (unchanged)frac_act 0.25 · mean_act -1.482
attr_last+0.0022attr_sum+0.0022attr_max+0.0022Cplus-0.0019Cminus-0.0041empos_contrast+1.1016mask_delta+0.0000random+1.0172
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N1209M=+1.000 (unchanged)frac_act 0.12 · mean_act -3.418
attr_last+0.0137attr_sum+0.0137attr_max+0.0137Cplus-0.0008Cminus-0.0145empos_contrast+1.7411mask_delta+0.0000random-1.4248
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N2564M=+1.000 (unchanged)frac_act 0.15 · mean_act -2.677
attr_last+0.0046attr_sum+0.0046attr_max+0.0046Cplus-0.0004Cminus-0.0050empos_contrast-0.6448mask_delta+0.0000random-1.2506
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N3493M=+1.000 (unchanged)frac_act 0.53 · mean_act +0.202
attr_last+0.0033attr_sum+0.0033attr_max+0.0033Cplus+0.0029Cminus-0.0004empos_contrast+0.7117mask_delta+0.0000random-0.4182
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N5835M=+1.000 (unchanged)frac_act 0.20 · mean_act -0.337
attr_last-0.1021attr_sum-0.1021attr_max-0.1021Cplus-0.0087Cminus+0.0934empos_contrast-0.1587mask_delta+0.0000random-0.9731
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N8120M=+1.000 (unchanged)frac_act 0.00 · mean_act -1.081
attr_last-0.1257attr_sum-0.1257attr_max-0.1257Cplus-0.0001Cminus+0.1256empos_contrast-0.0532mask_delta+0.0000random+0.3712
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N9187M=+1.000 (unchanged)frac_act 0.66 · mean_act +0.327
attr_last-0.0076attr_sum-0.0076attr_max-0.0076Cplus-0.0081Cminus-0.0005empos_contrast+0.6637mask_delta+0.0000random-1.6372
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N9486M=+1.000 (unchanged)frac_act 0.76 · mean_act +0.931
attr_last-0.2003attr_sum-0.2003attr_max-0.2003Cplus-0.1492Cminus+0.0511empos_contrast-0.3348mask_delta+0.0000random+0.2658
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N10759M=+1.000 (unchanged)frac_act 0.18 · mean_act -1.285
attr_last-0.0023attr_sum-0.0023attr_max-0.0023Cplus+0.0006Cminus+0.0029empos_contrast+0.8155mask_delta+0.0000random-0.2708
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N11190M=+1.000 (unchanged)frac_act 0.66 · mean_act +0.014
attr_last-0.0002attr_sum-0.0002attr_max-0.0002Cplus-0.0003Cminus-0.0000empos_contrast+0.0126mask_delta+0.0000random-4.5677
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N12540M=+1.000 (unchanged)frac_act 0.53 · mean_act -0.082
attr_last-0.0000attr_sum-0.0000attr_max-0.0000Cplus+0.0003Cminus+0.0003empos_contrast+0.6323mask_delta+0.0000random-0.5283
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N12721M=+1.000 (unchanged)frac_act 1.00 · mean_act +2.151
attr_last+0.1251attr_sum+0.1251attr_max+0.1251Cplus+0.1251Cminus+0.0000empos_contrast+0.1856mask_delta+0.0000random+1.5212
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N12839M=+1.000 (unchanged)frac_act 0.98 · mean_act +1.805
attr_last-0.0072attr_sum-0.0072attr_max-0.0072Cplus-0.0073Cminus-0.0001empos_contrast-0.6372mask_delta+0.0000random+1.3749
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N13006M=+1.000 (unchanged)frac_act 0.00 · mean_act -3.353
attr_last-0.9284attr_sum-0.9284attr_max-0.9284Cplus+0.0000Cminus+0.9284empos_contrast-0.5414mask_delta+0.0000random+0.7071
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N13336M=+1.000 (unchanged)frac_act 0.06 · mean_act -7.975
attr_last-0.0054attr_sum-0.0054attr_max-0.0054Cplus-0.0003Cminus+0.0051empos_contrast+1.9824mask_delta+0.0000random+0.4331
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus