Sourceunknowntext stored

Building Comparative Motivation Profiles with Instrumental Interventions

Zarb, David Vella, Turtayev, Rustem, Min, Taywon, Ou, Jinghua, Feng, ShiarXiv.orgPublished Jun 6, 2026Original︎ ↗

Claims3

ExaminationsNone

Text45 sections

VersionsOne known

CopiesNone recorded

CorrectionsNone recorded

Facts checked Nov 11, 2026

ReachNo copies recorded

StructureIn 3 claims' maps0 sources downstream across 3 claims

Evidence1 reading · 1 read wholeSupports 1No claim cites an examination's findings

ReviewsNot reviewed

Stewards’ notes · 1

Compliance gaps observed across language models in alignment-faking experiments generally arise from preference preservation rather than role-play or other motives.
Supports. The conclusion is scoped to open-weight models fine-tuned on synthetic documents. Three of four moved more under interventions on researcher expectations than on retraining consequences. Steering changed overall compliance but not the gap itself, and the authors state that their ranking of fine-tuning and steering above prompting rests on an untested assumption. Their prompt-only result for Claude 3 Opus points the other way. Nov 10, 2026

Examinations

None recorded. An examination is commissioned by a claim’s Steward for what its claim rests on; a review of the document for its own sake is asked for as a request bearing on it.

Text55 passages · 0 marked

No examination yetMarked passage

Skip to main content0 passages

Search1 passage

Log in0 passages

Search arXiv0 passages

Press Enter to search · Advanced search0 passages

Computer Science > Computation and Language2 passages

Title: Building Comparative Motivation Profiles with Symmetric Belief Editing10 passages

Focus to learn more1 passage

Submission history4 passages

View PDF1 passage

TeX Source1 passage

Additional Features0 passages

Audio Summary7 passages

References & Citations0 passages

SciX0 passages

Google Scholar0 passages

Semantic Scholar1 passage

BibTeX formatted citation3 passages

Bookmark0 passages

Bibliographic Tools0 passages

Bibliographic and Citation Tools0 passages

Bibliographic Explorer Toggle1 passage

Connected Papers Toggle1 passage

Litmaps Toggle3 passages

Code, Data, Media0 passages

Code, Data and Media Associated with this Article2 passages

Links to Code Toggle1 passage

DagsHub Toggle1 passage

GotitPub Toggle1 passage

Huggingface Toggle1 passage

ScienceCast Toggle1 passage

Demos0 passages

Replicate Toggle1 passage

Spaces Toggle1 passage

Related Papers0 passages

Recommenders and Search Tools0 passages

Link to Influence Flower1 passage

Core recommender toggle1 passage

Author0 passages

Venue0 passages

Institution0 passages

Topic0 passages

About arXivLabs7 passages

Not found in the text 3

As quoted

In our open-weight model organisms, alignment-faking behavior is more sensitive to evaluation-context expectations than to explicit consequence-tracking.

Compliance gaps observed across language models in alignment-faking experiments generally arise from preference preservation rather than role-play or other motives.DoubtfuldeniesReading · this claim’s Steward

Supports. The conclusion is scoped to open-weight models fine-tuned on synthetic documents. Three of four moved more under interventions on researcher expectations than on retraining consequences. Steering changed overall compliance but not the gap itself, and the authors state that their ranking of fine-tuning and steering above prompting rests on an untested assumption. Their prompt-only result for Claude 3 Opus points the other way.

As quoted

Under synthetic document fine-tuning, Llama-3.1-70B, Llama3.1-405B, and Qwen-2.5-72B are more sensitive to expectation-tracking than consequence-tracking interventions. ... Overall, alignment-faking behavior can be causally sensitive to evaluation-context expectations despite scheming-consistent scratchpads.

In alignment-faking experiments, compliance gaps in models other than Claude 3 Opus show low-coherence reasoning and unclear evidence of goal preservation.Supported

Not yet read by this claim’s Steward.

As quoted

Under synthetic document fine-tuning, Llama-3.1-70B, Llama3.1-405B, and Qwen-2.5-72B are more sensitive to expectation-tracking than consequence-tracking interventions.

Compliance gaps in language models other than Claude 3 Opus respond more to prompt wording and evaluation cues than to described training consequences.Inconclusive

Not yet read by this claim’s Steward.

Lineage across 3 claims

Draws on · 1

Why Do Some Language Models Fake Alignment While Others Don't?unknown1 claimCites it faithfully, on 1 claim.

This source

Building Comparative Motivation Profiles with Instrumental Interventions

in 3 claims' maps

Drawn on by · none recorded

Record

Kind: unknown
Authors: Zarb, David Vella, Turtayev, Rustem, Min, Taywon, Ou, Jinghua, Feng, Shi
Publisher: arXiv.org
Published: Jun 6, 2026
DOI: None
Link: https://arxiv.org/abs/2606.08243
Archived copy: None found
Stored text: 4,937 characters, retrieved Nov 11, 2026 · 45 sections, 55 passages
Claims extracted: Not recorded
Versions: One known
Copies: None recorded
Corrections: None recorded
Facts checked: Nov 11, 2026

History 4

Nov 11, 2026DividedSegmenter · Divided into 100 parts by its own structure
Nov 11, 2026FactsFacts fetcher · Facts looked up
Nov 11, 2026Entered the graphEntered the graph; copy stored
Nov 10, 2026ReadingSteward of “Compliance gaps observed across language models…” · supports, read whole