MétaCan
Menu
← Back to cohort
Record W4404305672 · doi:10.48550/arxiv.2407.18213

Scaling Trends in Language Model Robustness

2024· preprint· en· W4404305672 on OpenAlexfundno aff
Nikolaus H. R. Howe, Ian R. McKenzie, Oskar J. Hollinsworth, Michał Zając, Tom Tseng, Aaron Tucker, Pierre‐Luc Bacon, Adam Gleave

Bibliographic record

VenuearXiv (Cornell University) · 2024
Typepreprint
Languageen
FieldComputer Science
TopicNatural Language Processing Techniques
Canadian institutionsnot available
FundersNatural Sciences and Engineering Research Council of Canada
KeywordsRobustness (evolution)Computer scienceScale (ratio)Language modelNatural language processingGeographyCartographyChemistry

Abstract

fetched live from OpenAlex

Increasing model size has unlocked a dazzling array of capabilities in modern language models. At the same time, even frontier models remain vulnerable to jailbreaks and prompt injections, despite concerted efforts to make them robust. As both attack and defense gain access to more compute, and as models become larger, what happens to robustness? We argue that to answer this question requires a \emph{scaling} approach, which we employ in an extensive study of language model robustness across several classification tasks, model families, and adversarial attacks. We find that in the absence of explicit safety training, larger models are not consistently more robust; however, scale improves sample efficiency in adversarial training, though it worsens compute efficiency. Further, we find that increasing attack compute smoothly improves attack success rate against both undefended and adversarially trained models. Finally, after exploring robustness transfer across attacks and threat models, we combine attack and defense scaling rates to study the offense-defense balance. We find that while attack scaling outpaces adversarial training across all models studied, larger adversarially trained models might give defense the advantage in the long run. These results underscore the utility of the scaling lens, and provide a paradigm for evaluating future attacks and defenses on frontier models.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

How this classification was reachedexpand

Full frame machine prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.

metaresearch head score (Codex)0.006
metaresearch head score (Gemma)0.051
Version: metacan-v3-hybrid-931329e0061cValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Simulation or modeling · Consensus signal: Simulation or modeling
GenreCandidate signal: Empirical · Consensus signal: Empirical
Teacher disagreement score0.006
Threshold uncertainty score0.033

Distilled classifier scores by category (both heads)

CategoryCodexGemma
Metaresearch0.0060.051
Meta-epidemiology (narrow)0.0010.001
Meta-epidemiology (broad)0.0010.001
Bibliometrics0.0010.001
Science and technology studies0.0010.003
Scholarly communication0.0020.010
Open science0.0010.003
Research integrity0.0020.007
Insufficient payload (model declined to judge)0.0040.001

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.054
GPT teacher head0.224
Teacher spread0.170 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.

The models applied no category: nothing in the taxonomy fit this work.
Study designSimulation or modeling
Domainnot available
GenreEmpirical

How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".

Quick stats

Citations0
Published2024
Admission routes1
Has abstractyes

Explore more

Same venuearXiv (Cornell University)→Same topicNatural Language Processing Techniques→French-language works237,207→