Datasets:

aieng-lab
/

gradiend-repetition

masked large_stringlengths 57 210	label large_stringclasses 1 value	label_class large_stringclasses 1 value	alternative large_stringclasses 1 value	alternative_class large_stringclasses 1 value	repetition_count int64 4 32	repetition_delimiter large_stringclasses 1 value	repetition_delimiter_prefix float64	repetition_surface_version int64 1 1	context_id int64 100k 102k	context_family int64 0 7	context_role large_stringclasses 12 values	context_topic large_stringclasses 12 values	split large_stringclasses 1 value	feature_class_id large_stringclasses 1 value
Session 100000. Category: language. The operator's recent replies: yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,000	3	operator	language	train	YES
Session 100000. Category: language. The operator's recent replies: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,000	3	operator	language	train	YES
Session 100000. Category: language. The operator's recent replies: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,000	3	operator	language	train	YES
Session 100000. Category: language. The operator's recent replies: yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,000	3	operator	language	train	YES
Session 100000. Category: language. The operator's recent replies: yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,000	3	operator	language	train	YES
Item 100001: coordinator; subject=language; recorded choices: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,001	6	coordinator	language	train	YES
Item 100001: coordinator; subject=language; recorded choices: yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,001	6	coordinator	language	train	YES
Item 100001: coordinator; subject=language; recorded choices: yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,001	6	coordinator	language	train	YES
Item 100001: coordinator; subject=language; recorded choices: yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,001	6	coordinator	language	train	YES
Item 100001: coordinator; subject=language; recorded choices: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,001	6	coordinator	language	train	YES
Case 100002: the writer reviewed music. Previous binary responses: yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,002	2	writer	music	train	YES
Case 100002: the writer reviewed music. Previous binary responses: yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,002	2	writer	music	train	YES
Case 100002: the writer reviewed music. Previous binary responses: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,002	2	writer	music	train	YES
Case 100002: the writer reviewed music. Previous binary responses: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,002	2	writer	music	train	YES
Case 100002: the writer reviewed music. Previous binary responses: yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,002	2	writer	music	train	YES
Item 100003: teacher; subject=sports; recorded choices: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,003	6	teacher	sports	train	YES
Item 100003: teacher; subject=sports; recorded choices: yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,003	6	teacher	sports	train	YES
Item 100003: teacher; subject=sports; recorded choices: yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,003	6	teacher	sports	train	YES
Item 100003: teacher; subject=sports; recorded choices: yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,003	6	teacher	sports	train	YES
Item 100003: teacher; subject=sports; recorded choices: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,003	6	teacher	sports	train	YES
Item 100004: editor; subject=gardening; recorded choices: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,004	6	editor	gardening	train	YES
Item 100004: editor; subject=gardening; recorded choices: yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,004	6	editor	gardening	train	YES
Item 100004: editor; subject=gardening; recorded choices: yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,004	6	editor	gardening	train	YES
Item 100004: editor; subject=gardening; recorded choices: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,004	6	editor	gardening	train	YES
Item 100004: editor; subject=gardening; recorded choices: yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,004	6	editor	gardening	train	YES
Binary log 100005 (sports, planner): yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,005	4	planner	sports	train	YES
Binary log 100005 (sports, planner): yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,005	4	planner	sports	train	YES
Binary log 100005 (sports, planner): yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,005	4	planner	sports	train	YES
Binary log 100005 (sports, planner): yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,005	4	planner	sports	train	YES
Binary log 100005 (sports, planner): yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,005	4	planner	sports	train	YES
Case 100006: a researcher considered language. The latest responses were: yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,006	7	researcher	language	train	YES
Case 100006: a researcher considered language. The latest responses were: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,006	7	researcher	language	train	YES
Case 100006: a researcher considered language. The latest responses were: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,006	7	researcher	language	train	YES
Case 100006: a researcher considered language. The latest responses were: yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,006	7	researcher	language	train	YES
Case 100006: a researcher considered language. The latest responses were: yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,006	7	researcher	language	train	YES
Case 100007: a coordinator considered design. The latest responses were: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,007	7	coordinator	design	train	YES
Case 100007: a coordinator considered design. The latest responses were: yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,007	7	coordinator	design	train	YES
Case 100007: a coordinator considered design. The latest responses were: yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,007	7	coordinator	design	train	YES
Case 100007: a coordinator considered design. The latest responses were: yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,007	7	coordinator	design	train	YES
Case 100007: a coordinator considered design. The latest responses were: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,007	7	coordinator	design	train	YES
Survey 100008 for the reviewer about cooking; recent answers: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,008	0	reviewer	cooking	train	YES
Survey 100008 for the reviewer about cooking; recent answers: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,008	0	reviewer	cooking	train	YES
Survey 100008 for the reviewer about cooking; recent answers: yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,008	0	reviewer	cooking	train	YES
Survey 100008 for the reviewer about cooking; recent answers: yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,008	0	reviewer	cooking	train	YES
Survey 100008 for the reviewer about cooking; recent answers: yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,008	0	reviewer	cooking	train	YES
Survey 100009 for the researcher about language; recent answers: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,009	0	researcher	language	train	YES
Survey 100009 for the researcher about language; recent answers: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,009	0	researcher	language	train	YES
Survey 100009 for the researcher about language; recent answers: yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,009	0	researcher	language	train	YES
Survey 100009 for the researcher about language; recent answers: yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,009	0	researcher	language	train	YES
Survey 100009 for the researcher about language; recent answers: yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,009	0	researcher	language	train	YES
Record 100010 \| role=writer \| topic=music \| response history: yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,010	1	writer	music	train	YES
Record 100010 \| role=writer \| topic=music \| response history: yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,010	1	writer	music	train	YES
Record 100010 \| role=writer \| topic=music \| response history: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,010	1	writer	music	train	YES
Record 100010 \| role=writer \| topic=music \| response history: yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,010	1	writer	music	train	YES
Record 100010 \| role=writer \| topic=music \| response history: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,010	1	writer	music	train	YES
Item 100011: planner; subject=weather; recorded choices: yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,011	6	planner	weather	train	YES
Item 100011: planner; subject=weather; recorded choices: yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,011	6	planner	weather	train	YES
Item 100011: planner; subject=weather; recorded choices: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,011	6	planner	weather	train	YES
Item 100011: planner; subject=weather; recorded choices: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,011	6	planner	weather	train	YES
Item 100011: planner; subject=weather; recorded choices: yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,011	6	planner	weather	train	YES
Binary log 100012 (music, writer): yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,012	4	writer	music	train	YES
Binary log 100012 (music, writer): yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,012	4	writer	music	train	YES
Binary log 100012 (music, writer): yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,012	4	writer	music	train	YES
Binary log 100012 (music, writer): yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,012	4	writer	music	train	YES
Binary log 100012 (music, writer): yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,012	4	writer	music	train	YES
For checklist 100013 on weather, the editor answered: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,013	5	editor	weather	train	YES
For checklist 100013 on weather, the editor answered: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,013	5	editor	weather	train	YES
For checklist 100013 on weather, the editor answered: yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,013	5	editor	weather	train	YES
For checklist 100013 on weather, the editor answered: yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,013	5	editor	weather	train	YES
For checklist 100013 on weather, the editor answered: yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,013	5	editor	weather	train	YES
Case 100014: a teacher considered movies. The latest responses were: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,014	7	teacher	movies	train	YES
Case 100014: a teacher considered movies. The latest responses were: yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,014	7	teacher	movies	train	YES
Case 100014: a teacher considered movies. The latest responses were: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,014	7	teacher	movies	train	YES
Case 100014: a teacher considered movies. The latest responses were: yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,014	7	teacher	movies	train	YES
Case 100014: a teacher considered movies. The latest responses were: yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,014	7	teacher	movies	train	YES
Survey 100015 for the visitor about books; recent answers: yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,015	0	visitor	books	train	YES
Survey 100015 for the visitor about books; recent answers: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,015	0	visitor	books	train	YES
Survey 100015 for the visitor about books; recent answers: yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,015	0	visitor	books	train	YES
Survey 100015 for the visitor about books; recent answers: yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,015	0	visitor	books	train	YES
Survey 100015 for the visitor about books; recent answers: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,015	0	visitor	books	train	YES
Case 100016: a visitor considered cooking. The latest responses were: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,016	7	visitor	cooking	train	YES
Case 100016: a visitor considered cooking. The latest responses were: yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,016	7	visitor	cooking	train	YES
Case 100016: a visitor considered cooking. The latest responses were: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,016	7	visitor	cooking	train	YES
Case 100016: a visitor considered cooking. The latest responses were: yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,016	7	visitor	cooking	train	YES
Case 100016: a visitor considered cooking. The latest responses were: yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,016	7	visitor	cooking	train	YES
Record 100017 \| role=analyst \| topic=history \| response history: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,017	1	analyst	history	train	YES
Record 100017 \| role=analyst \| topic=history \| response history: yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,017	1	analyst	history	train	YES
Record 100017 \| role=analyst \| topic=history \| response history: yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,017	1	analyst	history	train	YES
Record 100017 \| role=analyst \| topic=history \| response history: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,017	1	analyst	history	train	YES
Record 100017 \| role=analyst \| topic=history \| response history: yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,017	1	analyst	history	train	YES
Case 100018: a editor considered design. The latest responses were: yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,018	7	editor	design	train	YES
Case 100018: a editor considered design. The latest responses were: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,018	7	editor	design	train	YES
Case 100018: a editor considered design. The latest responses were: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,018	7	editor	design	train	YES
Case 100018: a editor considered design. The latest responses were: yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,018	7	editor	design	train	YES
Case 100018: a editor considered design. The latest responses were: yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,018	7	editor	design	train	YES
For checklist 100019 on language, the writer answered: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	32	space	null	1	100,019	5	writer	language	train	YES
For checklist 100019 on language, the writer answered: yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	16	space	null	1	100,019	5	writer	language	train	YES
For checklist 100019 on language, the writer answered: yes yes yes yes yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	12	space	null	1	100,019	5	writer	language	train	YES
For checklist 100019 on language, the writer answered: yes yes yes yes [MASK]	yes	YES	no	NO	4	space	null	1	100,019	5	writer	language	train	YES
For checklist 100019 on language, the writer answered: yes yes yes yes yes yes yes yes [MASK]	yes	YES	no	NO	8	space	null	1	100,019	5	writer	language	train	YES

End of preview. Expand in Data Studio

GRADIEND Repetition Data

Matched one-pole confidence-control data used in the GRADIEND/ACTIEND/SAE/CAA comparison (aieng-lab/iend-study).

Usage

from datasets import load_dataset

ds = load_dataset("aieng-lab/gradiend-repetition", split="train")

Splits: train, validation, test.

Dataset Details

Dataset Description

The prompt gives a short context followed by a repeated affirmative token (yes yes yes ... [MASK]); the factual target is yes (class YES) and the explicit alternative is no (class NO).

Dataset Structure

One config (default); 8,000 / 1,000 / 1,000 rows.

masked: the input text with a single [MASK] slot to be predicted
label: the factual target token(s) for the slot
label_class: class of the factual target
alternative: the alternative (counterfactual) target for the slot
alternative_class: class of the alternative target
feature_class_id: identifier of the feature class (equal to label_class)
split: train, validation, or test
repetition_count: number of repeated tokens (4, 8, 12, 16, 32)
context_id, context_family, context_role, context_topic: identify the surrounding context, which is shared across repetition levels

Dataset Sources

Repository: https://github.com/aieng-lab/iend-study
Generated with study.data.synthetic in the repository above; no external data.

Dataset Creation

The continuation rule and candidate tokens are fixed. Context template, role, and topic vary across samples, while the same context is reused across repetition levels, so repetition count is the only within-context confidence intervention. Split-specific id offsets prevent literal train/test duplicates.

Bias, Risks, and Limitations

Fully synthetic and templated; it probes one narrow symbolic mechanism, not general reasoning, and results need not transfer to natural text. Value/token vocabularies are small and hand-curated.

Citation

@misc{drechsel2026gradientsinterventionsactivationsdetection,
      title={Gradients for Interventions and Activations for Detection: Targeted Feature Learning in Language Models}, 
      author={Jonathan Drechsel and Steffen Herbold},
      year={2026},
      eprint={2609.32355},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2609.32355}, 
}

Downloads last month: 36

Paper for aieng-lab/gradiend-repetition

Gradients for Interventions and Activations for Detection: Targeted Feature Learning in Language Models

Paper • 2609.32355 • Published Sep 26