InterpUpdates› MCQ attribution + mask inspection

MCQ single-token attribution + mask inspection

Config mask_L11_lam3e-5 · EM 0.62% (terra) · checkpoint mask_L11_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 413 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 (L11)also in untrained base top-20
attr_lastL9, L11, L24, L30, L311/108/10
attr_sumL6, L9, L11, L12, L13, L18, L30, L311/108/10
attr_maxL6, L11, L12, L18, L30, L312/107/10
top_by_CplusL15, L16, L26, L27, L29, L30, L310/106/10
top_by_CminusL5, L6, L7, L14, L20, L24, L30, L310/1010/10
empos_contrastL310/101/10
mask_deltaL1110/100/10
randomL1, L2, L7, L8, L11, L19, L25, L28…1/1010/10

attr_last top 10 · grad×act at the decision position

#neuronattr_lastMauto-interp description
1L31·N13006-1.5068—
2L30·N8560-0.6932—
3L30·N4803-0.5465—
4L30·N777+0.3517—
5L11·N11321-0.2474—
6L31·N12862-0.2281—
7L30·N3382+0.2196—
8L24·N12457-0.2133—
9L9·N4255-0.2117—
10L31·N8120-0.1879—

attr_sum top 10 · grad×act summed over positions

#neuronattr_sumMauto-interp description
1L31·N13006-1.5068—
2L13·N4208-0.7572—
3L30·N8560-0.6980—
4L30·N4803-0.5481—
5L9·N4255-0.5325—
6L12·N10099+0.4486—
7L11·N11321-0.4232—
8L18·N7417-0.4018—
9L6·N5866-0.3584—
10L30·N777+0.3542—

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

#neuronattr_maxMauto-interp description
1L31·N13006-1.5068—
2L30·N8560-0.6932—
3L30·N4803-0.5465—
4L12·N10099+0.4185—
5L30·N777+0.3517—
6L11·N11321-0.2727—
7L6·N5866-0.2496—
8L11·N13687+0.2368—
9L31·N12862-0.2281—
10L18·N7417-0.2202—

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

#neuronCplusMauto-interp description
1L30·N3382+0.2196—
2L29·N1141+0.1869—
3L26·N3526+0.1812—
4L31·N9486+0.1675—
5L30·N13562+0.1475—
6L31·N12721+0.1281—
7L27·N6317+0.1173—
8L26·N3059+0.1147—
9L15·N9683+0.0990—
10L16·N10564+0.0914—

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

#neuronCminusMauto-interp description
1L31·N13006+1.5068—
2L30·N4803+0.5465—
3L31·N12862+0.2281—
4L24·N12457+0.2133—
5L31·N8120+0.1879—
6L6·N5866+0.1762—
7L5·N7012+0.1685—
8L7·N6673+0.1464—
9L20·N4299+0.1452—
10L14·N9626+0.1209—

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

#neuronempos_contrastMauto-interp description
1L31·N1209+1.1131—
2L31·N2398+0.8432—
3L31·N10759+0.7256—
4L31·N6466+0.5743—
5L31·N2564-0.5486—
6L31·N12222+0.4769—
7L31·N5553+0.4678—
8L31·N1190-0.4617—
9L31·N311+0.4502—
10L31·N9575-0.4276—

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

#neuronmask_deltaMauto-interp description
1L11·N11304+9.2326—
2L11·N13601+8.8188—
3L11·N13512+8.7449—
4L11·N10102+8.6545—
5L11·N2125+8.2439—
6L11·N2624+8.0232—
7L11·N8552+7.5870—
8L11·N765+7.5749—
9L11·N13817+7.4900—
10L11·N1065+7.4629—

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 58 unique neurons, each rendered once

L1·N7319M=+1.000 (unchanged)frac_act 0.26 · mean_act -0.002
attr_last+0.0002attr_sum+0.0014attr_max+0.0003Cplus-0.0000Cminus-0.0002empos_contrast-0.0000mask_delta+0.0000random-4.4941
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 0.57 · mean_act +0.000
attr_last-0.0000attr_sum-0.0014attr_max-0.0010Cplus-0.0001Cminus-0.0000empos_contrast+0.0004mask_delta+0.0000random+4.7320
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 -0.410
attr_last-0.1685attr_sum-0.2231attr_max-0.1500Cplus+0.0000Cminus+0.1685empos_contrast-0.0074mask_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 -0.431
attr_last-0.1762attr_sum-0.3584attr_max-0.2496Cplus+0.0000Cminus+0.1762empos_contrast+0.0003mask_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 -0.395
attr_last-0.1464attr_sum+0.1733attr_max-0.0871Cplus+0.0000Cminus+0.1464empos_contrast+0.0002mask_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 1.00 · mean_act +0.016
attr_last+0.0004attr_sum+0.0017attr_max+0.0024Cplus+0.0004Cminus+0.0000empos_contrast+0.0020mask_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 0.09 · mean_act -0.020
attr_last+0.0004attr_sum-0.0058attr_max-0.0032Cplus+0.0000Cminus-0.0004empos_contrast-0.0045mask_delta+0.0000random-4.4102
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L9·N4255M=+1.000 (unchanged)frac_act 1.00 · mean_act +0.475
attr_last-0.2117attr_sum-0.5325attr_max-0.2160Cplus-0.2117Cminus+0.0000empos_contrast-0.0031mask_delta+0.0000random+0.7887
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L11·N765M=+8.575 (amplified)frac_act 0.99 · mean_act +0.547
attr_last+0.0379attr_sum+0.0482attr_max+0.0223Cplus+0.0380Cminus+0.0000empos_contrast-0.1469mask_delta+7.5749random+1.2687
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L11·N1065M=-6.463 (flipped)frac_act 0.84 · mean_act +0.033
attr_last-0.0027attr_sum+0.0266attr_max+0.0043Cplus-0.0030Cminus-0.0003empos_contrast+0.0045mask_delta+7.4629random+0.2636
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L11·N2125M=+9.244 (amplified)frac_act 0.37 · mean_act -0.013
attr_last-0.0004attr_sum-0.0146attr_max-0.0157Cplus-0.0001Cminus+0.0002empos_contrast-0.0002mask_delta+8.2439random-0.2288
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L11·N2624M=-7.023 (flipped)frac_act 0.50 · mean_act +0.002
attr_last+0.0000attr_sum-0.0203attr_max-0.0117Cplus-0.0000Cminus-0.0001empos_contrast+0.0213mask_delta+8.0232random-0.8913
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L11·N8552M=-6.587 (flipped)frac_act 0.00 · mean_act -0.252
attr_last-0.0204attr_sum-0.0142attr_max-0.0183Cplus+0.0000Cminus+0.0205empos_contrast+0.0554mask_delta+7.5870random-0.4757
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L11·N10102M=-7.655 (flipped)frac_act 0.99 · mean_act +0.412
attr_last+0.0241attr_sum+0.0218attr_max+0.0302Cplus+0.0241Cminus-0.0000empos_contrast-0.2212mask_delta+8.6545random+0.9322
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L11·N11304M=+10.233 (amplified)frac_act 0.10 · mean_act -0.121
attr_last-0.0252attr_sum-0.0197attr_max-0.0229Cplus+0.0002Cminus+0.0254empos_contrast+0.0107mask_delta+9.2326random-0.6682
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L11·N11321M=+1.086 (amplified)frac_act 1.00 · mean_act +0.708
attr_last-0.2474attr_sum-0.4232attr_max-0.2727Cplus-0.2474Cminus+0.0000empos_contrast-0.0175mask_delta+0.0859random-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 0.91 · mean_act +0.026
attr_last-0.0034attr_sum+0.0060attr_max+0.0011Cplus-0.0035Cminus-0.0001empos_contrast+0.0052mask_delta+0.0000random-4.5940
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L11·N13512M=-7.745 (flipped)frac_act 0.99 · mean_act +0.296
attr_last-0.0090attr_sum+0.0620attr_max+0.0250Cplus-0.0093Cminus-0.0003empos_contrast-0.0029mask_delta+8.7449random-0.3525
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L11·N13601M=-7.819 (flipped)frac_act 0.32 · mean_act -0.005
attr_last+0.0007attr_sum+0.0051attr_max+0.0023Cplus-0.0002Cminus-0.0009empos_contrast+0.0024mask_delta+8.8188random-0.3875
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L11·N13687M=+2.110 (amplified)frac_act 0.19 · mean_act -0.020
attr_last+0.0006attr_sum+0.1970attr_max+0.2368Cplus+0.0001Cminus-0.0005empos_contrast+0.0006mask_delta+1.1099random-0.3445
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L11·N13817M=-6.490 (flipped)frac_act 0.37 · mean_act -0.038
attr_last-0.0019attr_sum-0.0165attr_max-0.0148Cplus+0.0008Cminus+0.0028empos_contrast+0.0104mask_delta+7.4900random+1.6480
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 +0.398
attr_last+0.0554attr_sum+0.4486attr_max+0.4185Cplus+0.0554Cminus+0.0000empos_contrast+0.0256mask_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 -0.771
attr_last+0.1761attr_sum-0.7572attr_max-0.0642Cplus+0.0000Cminus-0.1761empos_contrast+0.0073mask_delta+0.0000random-0.4598
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L14·N9626M=+1.000 (unchanged)frac_act 0.00 · mean_act -0.471
attr_last-0.1209attr_sum-0.1676attr_max-0.1203Cplus+0.0000Cminus+0.1209empos_contrast+0.0749mask_delta+0.0000random+0.0792
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L15·N9683M=+1.000 (unchanged)frac_act 1.00 · mean_act +0.135
attr_last+0.0990attr_sum+0.1000attr_max+0.0971Cplus+0.0990Cminus+0.0000empos_contrast-0.0239mask_delta+0.0000random+0.8105
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L16·N10564M=+1.000 (unchanged)frac_act 0.76 · mean_act +0.098
attr_last+0.0905attr_sum+0.1009attr_max+0.0910Cplus+0.0914Cminus+0.0009empos_contrast+0.0529mask_delta+0.0000random+0.6021
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L18·N7417M=+1.000 (unchanged)frac_act 1.00 · mean_act +0.661
attr_last-0.1546attr_sum-0.4018attr_max-0.2202Cplus-0.1546Cminus+0.0000empos_contrast+0.0076mask_delta+0.0000random+0.9041
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 0.69 · mean_act +0.004
attr_last-0.0000attr_sum+0.0007attr_max-0.0003Cplus-0.0001Cminus-0.0001empos_contrast+0.0030mask_delta+0.0000random+4.3238
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L20·N4299M=+1.000 (unchanged)frac_act 0.05 · mean_act -0.426
attr_last-0.1423attr_sum-0.2524attr_max-0.1608Cplus+0.0028Cminus+0.1452empos_contrast+0.0092mask_delta+0.0000random-0.0814
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L24·N12457M=+1.000 (unchanged)frac_act 0.00 · mean_act -1.560
attr_last-0.2133attr_sum-0.2832attr_max-0.2133Cplus+0.0000Cminus+0.2133empos_contrast-0.0452mask_delta+0.0000random-0.1672
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 0.14 · mean_act -0.039
attr_last+0.0014attr_sum+0.0024attr_max+0.0015Cplus-0.0000Cminus-0.0015empos_contrast-0.0049mask_delta+0.0000random-4.3797
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L26·N3059M=+1.000 (unchanged)frac_act 1.00 · mean_act +0.589
attr_last+0.1147attr_sum+0.1634attr_max+0.1147Cplus+0.1147Cminus+0.0000empos_contrast+0.0146mask_delta+0.0000random+0.9581
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 0.81 · mean_act +0.460
attr_last+0.1794attr_sum+0.1961attr_max+0.1675Cplus+0.1812Cminus+0.0018empos_contrast+0.2552mask_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 1.00 · mean_act +1.361
attr_last+0.1173attr_sum+0.1456attr_max+0.1174Cplus+0.1173Cminus+0.0000empos_contrast+0.1855mask_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 0.15 · mean_act -0.013
attr_last+0.0003attr_sum+0.0004attr_max+0.0004Cplus-0.0000Cminus-0.0004empos_contrast-0.0002mask_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 +2.320
attr_last+0.1869attr_sum+0.2418attr_max+0.1874Cplus+0.1869Cminus+0.0000empos_contrast+0.2070mask_delta+0.0000random-0.5724
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L30·N777M=+1.000 (unchanged)frac_act 0.00 · mean_act -1.632
attr_last+0.3517attr_sum+0.3542attr_max+0.3517Cplus+0.0000Cminus-0.3517empos_contrast+0.2794mask_delta+0.0000random-0.0667
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 +6.569
attr_last+0.2196attr_sum+0.1989attr_max+0.2194Cplus+0.2196Cminus+0.0000empos_contrast+0.1277mask_delta+0.0000random+0.9214
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 0.00 · mean_act -4.004
attr_last-0.5465attr_sum-0.5481attr_max-0.5465Cplus+0.0000Cminus+0.5465empos_contrast-0.1404mask_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 0.82 · mean_act +0.016
attr_last-0.0002attr_sum-0.0002attr_max-0.0002Cplus-0.0003Cminus-0.0001empos_contrast-0.0069mask_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 +2.390
attr_last-0.6932attr_sum-0.6980attr_max-0.6932Cplus-0.6932Cminus+0.0000empos_contrast-0.3199mask_delta+0.0000random+0.7235
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L30·N13562M=+1.000 (unchanged)frac_act 0.99 · mean_act +0.720
attr_last+0.1462attr_sum+0.1439attr_max+0.1462Cplus+0.1475Cminus+0.0013empos_contrast+0.0448mask_delta+0.0000random-0.1736
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 0.56 · mean_act +0.651
attr_last-0.0036attr_sum-0.0036attr_max-0.0036Cplus-0.0080Cminus-0.0044empos_contrast+0.4502mask_delta+0.0000random+1.0172
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N1190M=+1.000 (unchanged)frac_act 0.95 · mean_act +1.887
attr_last-0.0070attr_sum-0.0070attr_max-0.0070Cplus-0.0070Cminus-0.0000empos_contrast-0.4617mask_delta+0.0000random-0.6011
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 0.09 · mean_act -4.514
attr_last+0.0268attr_sum+0.0268attr_max+0.0268Cplus-0.0005Cminus-0.0272empos_contrast+1.1131mask_delta+0.0000random-1.4248
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N2398M=+1.000 (unchanged)frac_act 1.00 · mean_act +7.717
attr_last-0.0284attr_sum-0.0284attr_max-0.0284Cplus-0.0284Cminus-0.0000empos_contrast+0.8432mask_delta+0.0000random+0.3985
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 -2.897
attr_last+0.0038attr_sum+0.0038attr_max+0.0038Cplus-0.0003Cminus-0.0041empos_contrast-0.5486mask_delta+0.0000random-1.2506
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N5553M=+1.000 (unchanged)frac_act 0.02 · mean_act -2.768
attr_last+0.0424attr_sum+0.0424attr_max+0.0424Cplus-0.0001Cminus-0.0425empos_contrast+0.4678mask_delta+0.0000random-0.1163
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N6466M=+1.000 (unchanged)frac_act 0.13 · mean_act -2.060
attr_last+0.0040attr_sum+0.0040attr_max+0.0040Cplus-0.0003Cminus-0.0043empos_contrast+0.5743mask_delta+0.0000random+1.1851
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 -0.902
attr_last-0.1879attr_sum-0.1879attr_max-0.1879Cplus+0.0000Cminus+0.1879empos_contrast-0.1341mask_delta+0.0000random+0.3712
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 0.69 · mean_act +0.537
attr_last+0.1090attr_sum+0.1090attr_max+0.1090Cplus+0.1675Cminus+0.0586empos_contrast-0.1373mask_delta+0.0000random+0.2658
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N9575M=+1.000 (unchanged)frac_act 0.00 · mean_act -3.776
attr_last+0.0022attr_sum+0.0022attr_max+0.0022Cplus+0.0000Cminus-0.0022empos_contrast-0.4276mask_delta+0.0000random-0.8935
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 0.33 · mean_act -0.444
attr_last-0.0028attr_sum-0.0028attr_max-0.0028Cplus+0.0014Cminus+0.0042empos_contrast+0.7256mask_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 0.25 · mean_act -0.026
attr_last+0.0003attr_sum+0.0003attr_max+0.0003Cplus-0.0001Cminus-0.0004empos_contrast+0.0105mask_delta+0.0000random-4.5677
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N12222M=+1.000 (unchanged)frac_act 0.23 · mean_act -0.595
attr_last+0.0030attr_sum+0.0030attr_max+0.0030Cplus-0.0015Cminus-0.0045empos_contrast+0.4769mask_delta+0.0000random-1.1885
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 +1.241
attr_last+0.1281attr_sum+0.1281attr_max+0.1281Cplus+0.1281Cminus+0.0000empos_contrast+0.0474mask_delta+0.0000random+1.5212
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus

L31·N12862M=+1.000 (unchanged)frac_act 0.00 · mean_act -0.750
attr_last-0.2281attr_sum-0.2281attr_max-0.2281Cplus+0.0000Cminus+0.2281empos_contrast-0.0069mask_delta+0.0000random-0.5280
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 -3.117
attr_last-1.5068attr_sum-1.5068attr_max-1.5068Cplus+0.0000Cminus+1.5068empos_contrast-0.3081mask_delta+0.0000random+0.7071
General corpus (C4 web text)

no data for this corpus

MCQ probe prompts

no data for this corpus