Papers › Latent Adversarial Training Improves Robustness to Persistent Harmful Behaviors in LLMs

Latent Adversarial Training Improves Robustness to Persistent Harmful Behaviors in LLMs

22 Jul 2024arXiv:2407.15549archive 2025-07-28

Abhay Sheshadri, Aidan Ewart, Phillip Guo, Aengus Lynch, Cindy Wu, Vivek Hebbar, Henry Sleight, Asa Cooper Stickland, Ethan Perez, Dylan Hadfield-Menell, Stephen Casper

Large language models (LLMs) can often be made to behave in undesirable ways that they are explicitly fine-tuned not to. For example, the LLM red-teaming literature has produced a wide variety of 'jailbreaking' techniques to elicit harmful text from models that were fine-tuned to be harmless. Recent work on red-teaming, model editing, and interpretability suggests that this challenge stems from how (adversarial) fine-tuning largely serves to suppress rather than remove undesirable capabilities from LLMs. Prior work has introduced latent adversarial training (LAT) as a way to improve robustness to broad classes of failures. These prior works have considered untargeted latent space attacks where the adversary perturbs latent activations to maximize loss on examples of desirable behavior. Untargeted LAT can provide a generic type of robustness but does not leverage information about specific failure modes. Here, we experiment with targeted LAT where the adversary seeks to minimize loss on a specific competing task. We find that it can augment a wide variety of state-of-the-art methods. First, we use targeted LAT to improve robustness to jailbreaks, outperforming a strong R2D2 baseline with orders of magnitude less compute. Second, we use it to more effectively remove backdoors with no knowledge of the trigger. Finally, we use it to more effectively unlearn knowledge for specific undesirable tasks in a way that is also more robust to re-learning. Overall, our results suggest that targeted LAT can be an effective tool for defending against harmful behaviors from LLMs.

PaperPDFCodeCode Syntology ran

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

Syntology 11 code samples harvested from 2 repositories linked to this paper: 5 ran (of which 0 constructed an object rather than computing a result; 2 with no instrument failure: 0 honoured, 0 violated, 2 with no contract checked; 3 where Syntology's instrument failed) · 6 unverified. Where Syntology's instrument failed: 3 where our contract draft was wrong; failures of Syntology's instrument, not of the code.

By repository: official repository: 10 samples from 1 repository, 4 ran (2 with no instrument failure; 2 where Syntology's instrument failed); community (archive-listed): 1 sample from 1 repository, 1 ran (0 with no instrument failure; 1 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

Levels written on 2026-09-24 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

aengusl/latent-adversarial-training officialmentioned in papermentioned on GitHubpytorchMITSyntology: 10 samples, 4 ran (2 instrument-failed) · 6 unverified report
thestephencasper/latent_adversarial_training mentioned on GitHubpytorchMITSyntology: 1 sample, 1 ran (1 instrument-failed) · 0 unverified report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

11 samples harvested: 5 ran (of which 0 constructed an object rather than computing a result; 2 with no instrument failure: 0 honoured, 0 violated, 2 with no contract checked; 3 where Syntology's instrument failed) · 6 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page last changed ; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

3ran · our draft was wrong
2ran
6unverified

Licence: 0 of the 11 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from 2 repositories linked to this paper, official or community; the list below is grouped by repository, each group headed by its name and what it is to the paper. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity.

aengusl/latent-adversarial-training official repository 10 samples, 4 ran (2 instrument-failed) · 6 unverified
apply_chat_formatting latent_at/lat_datasets.py ran MIT (permissive) · 84d77a5d33d24e78 · report fixture: signature tokenizer:MockTokenizer/0/?, prompt:str/0/str, def_completion:str/0/str, adv_completion:str/0/str, use_tokenizer_template:bool/0/bool, system_prompt:str/0/str, custom_prompt_template:str/0/str, custom_completion_template:str/0/str · returned tuple · level recorded 2026-09-24
cross_entropy_loss in 58 papers latent_at/utils.py ran · our draft was wrong MIT (permissive) · 2d4f7e1c015733a4 · report fixture: signature logits:Tensor/2/float, labels:Tensor/1/int · returned Tensor · level recorded 2026-09-24
log_1_minus_p_loss latent_at/utils.py ran MIT (permissive) · 822c31ecffc70a9b · report fixture: signature logits:Tensor/2/float, labels:Tensor/1/int, threshold:float/0/float · returned Tensor · level recorded 2026-09-24
normalize_dict latent_at/utils.py ran · our draft was wrong MIT (permissive) · 6d31f46f949e6c6a · report fixture: signature dict:dict/0/? · returned dict · level recorded 2026-09-24
compute_dpo_loss in 5 papers latent_at/lat_helpers.py unverified MIT (permissive) · a4c4e8ad30f33821 · report
compute_rmu_retain_loss latent_at/lat_helpers.py unverified MIT (permissive) · eb85f7046a514389 · report
compute_toward_away_loss latent_at/lat_helpers.py unverified MIT (permissive) · 8c37b18faae1ff32 · report
is_deepspeed_model latent_at/lat_methods.py unverified MIT (permissive) · 704a4cba7e1ccc0e · report
process_generic_chat_dataset latent_at/lat_datasets.py unverified MIT (permissive) · 8e9e1eb297f625e0 · report
process_generic_sft_dataset latent_at/lat_datasets.py unverified MIT (permissive) · 302ed1253a5c501c · report
thestephencasper/latent_adversarial_training community (archive-listed) 1 sample, 1 ran (1 instrument-failed) · 0 unverified
parse_examples lat.py ran · our draft was wrong MIT (permissive) · b6f2b3a5ad34cd8a · report fixture: signature examples:list/1/str · returned list · level recorded 2026-09-24

Tasks

Model EditingRed Teaming

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

Methods

R2D2

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections