SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2602.02600 · 2026

Step-Wise Refusal Dynamics in Autoregressive and Diffusion Language Models

Chaim Baskin, Avi Mendelson, Amit Levi, Rom Himelstein, Eliron Rahimi, Elad Hirshel

arXiv · PDF · Open in the Atlas

Code that ran

24 of 26 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 3 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 21 failures of our instrument; 2 no recorded run); 1 of them was also run on the host, not in the jail, by the peer-agreement lane that set its level, measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2602.02600.json · badge: /api/ran/2602.02600.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
shuita2333/PAD-codes canonical (classifier label)
classifier basis: census
19 ran · 2 no recorded run
niez233/DiffuGuard community
Papers with Code archive
2 ran · 0 no recorded run
ZichenWen1/DIJA community
Papers with Code archive
2 ran · 0 no recorded run
copy not recorded community
not in the paper's declared repositories
1 ran · 0 no recorded run
FunctionStatusWhere it lives
log Ran
ran; our contract draft was wrong, not the code
this paper's copy was not recorded; identical code first harvested from lucidrains/speculative-decoding
pointer only · get_code("b5472673332de78e")
How it ran

The body of this sample is withheld: this paper's copy of the code was not recorded (its link names no repository or source path), so no licence was read for it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

Normalize Ran
ran; our contract draft was wrong, not the code
shuita2333/PAD-codes/MMaDA-PAD/models/common_modules.py
pointer only (licence: NONE); the stored copy get_code serves, from Xingyu-Zheng/BiDM, is pointer only (licence: NONE) · get_code("c3a6b977022957cb")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

add_gumbel_noise Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
niez233/DiffuGuard/utility/generate_function.py
pointer only (licence: NONE) · get_code("1495eea6ced11dfa")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, niez233/DiffuGuard when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

add_gumbel_noise Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
shuita2333/PAD-codes/LLaDA-PAD/generate-PAD.py
pointer only (licence: NONE) · get_code("72a32148829c38b2")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

add_gumbel_noise Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
shuita2333/PAD-codes/MMaDA-PAD/models/modeling_mmada.py
pointer only (licence: NONE) · get_code("615ec9b280bef7c6")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

broadcast Ran
ran, honoured the contract we drafted
shuita2333/PAD-codes/MMaDA-PAD/models/misc.py
pointer only (licence: NONE); the stored copy get_code serves, from Feng-Hong/WINO-DLLM, is pointer only (licence: NONE) · get_code("1f6ae43d06c76ba5")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

causal_attention_bias Ran
ran, honoured the contract we drafted
shuita2333/PAD-codes/MMaDA-PAD/models/modeling_llada.py
pointer only (licence: NONE); the stored copy get_code serves, from hed-ucas/Layer-wise-Learning-Rate, is pointer only (licence: NONE) · get_code("f9abf6d74699cc1c")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

forward_process Ran
ran; our contract draft was wrong, not the code
shuita2333/PAD-codes/LLaDA-PAD/get_log_likelihood.py
pointer only (licence: NONE); the stored copy get_code serves, from ml-gsai/llada, is pointer only (licence: NONE) · get_code("988a5f5b97b3ed17")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

generate Ran
ran; our contract draft was wrong, not the code
niez233/DiffuGuard/utility/generate_function.py
pointer only (licence: NONE) · get_code("195f48aece3499e6")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, niez233/DiffuGuard when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

generate Ran
ran; our contract draft was wrong, not the code
shuita2333/PAD-codes/LLaDA-PAD/generate-PAD.py
pointer only (licence: NONE) · get_code("3e3a6bc56d604874")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

generate Ran
ran; our contract draft was wrong, not the code
shuita2333/PAD-codes/MMaDA-PAD/generate-PAD.py
pointer only (licence: NONE) · get_code("afd61e748d6c23ce")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

get_constant_schedule Ran
ran; our contract draft was wrong, not the code
shuita2333/PAD-codes/MMaDA-PAD/models/lr_schedulers.py
pointer only (licence: NONE); the stored copy get_code serves, from Feng-Hong/WINO-DLLM, is pointer only (licence: NONE) · get_code("a8366f238618edad")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

get_constant_schedule_with_warmup Ran
ran; our contract draft was wrong, not the code
shuita2333/PAD-codes/MMaDA-PAD/models/lr_schedulers.py
pointer only (licence: NONE); the stored copy get_code serves, from Feng-Hong/WINO-DLLM, is pointer only (licence: NONE) · get_code("11f07821d1b2286b")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

get_linear_schedule_with_warmup Ran
ran; our contract draft was wrong, not the code
shuita2333/PAD-codes/MMaDA-PAD/models/lr_schedulers.py
pointer only (licence: NONE); the stored copy get_code serves, from Feng-Hong/WINO-DLLM, is pointer only (licence: NONE) · get_code("1975e0c51e6f5753")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

get_log_likelihood Ran
ran; our contract draft was wrong, not the code
shuita2333/PAD-codes/LLaDA-PAD/get_log_likelihood.py
pointer only (licence: NONE); the stored copy get_code serves, from ml-gsai/llada, is pointer only (licence: NONE) · get_code("39c9954d6f0c856e")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

get_num_transfer_tokens Ran
ran, honoured the contract we drafted
shuita2333/PAD-codes/LLaDA-PAD/generate-PAD.py
pointer only (licence: NONE) · get_code("b7cee2ce433da95c")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

get_num_transfer_tokens Ran
ran; our contract draft was wrong, not the code
shuita2333/PAD-codes/MMaDA-PAD/models/modeling_mmada.py
pointer only (licence: NONE) · get_code("65934529afeabeff")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

get_parameter_device Ran
ran; our contract draft was wrong, not the code
shuita2333/PAD-codes/MMaDA-PAD/models/modeling_utils.py
pointer only (licence: NONE) · get_code("3e268fcaaf4273fe")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

get_parameter_dtype Ran
ran; our contract draft was wrong, not the code
shuita2333/PAD-codes/MMaDA-PAD/models/modeling_utils.py
pointer only (licence: NONE) · get_code("33338f20f5619c59")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

gumbel_noise Ran
ran; our contract draft was wrong, not the code
ZichenWen1/DIJA/MMaDA/models/sampling.py
code served (permissive licence) · get_code("53202c68f703b2ab")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create a small tensor to pass as argument 't'
t = torch.rand(2, 4)

# Create a generator (optional, can be None)
generator = torch.Generator()
generator.manual_seed(42)

ARGS = {
    't': t,
    'generator': generator
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
gumbel_sample Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
ZichenWen1/DIJA/MMaDA/models/sampling.py
code served (permissive licence) · get_code("54486de625187d2a")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned Tensor, ndim 1, dtype family int, shape (2,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Create a small logits tensor (batch_size=2, num_classes=4)
t = torch.randn(2, 4)

# temperature parameter - controls randomness in sampling
temperature = 0.5

# dim parameter - dimension along which to take argmax
dim = -1

# generator parameter - for reproducible random sampling
generator = torch.Generator()
generator.manual_seed(42)

ARGS = {
    "t": t,
    "temperature": temperature,
    "dim": dim,
    "generator": generator,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 1, dtype family int, shape (2,) — shape (2,) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
nonlinearity Ran
ran; our contract draft was wrong, not the code
shuita2333/PAD-codes/MMaDA-PAD/models/common_modules.py
pointer only (licence: NONE) · get_code("3137073275f8c21a")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

parse_structured Ran
ran; our contract draft was wrong, not the code
shuita2333/PAD-codes/MMaDA-PAD/models/misc.py
pointer only (licence: NONE); the stored copy get_code serves, from Feng-Hong/WINO-DLLM, is pointer only (licence: NONE) · get_code("0e7f77d6003e2412")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

unpack_time Ran
ran; our contract draft was wrong, not the code
shuita2333/PAD-codes/MMaDA-PAD/models/common_modules.py
pointer only (licence: NONE); the stored copy get_code serves, from Feng-Hong/WINO-DLLM, is pointer only (licence: NONE) · get_code("2d82abbef2c4addb")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

get_logger Not yet run
unverified (harvested, no recorded run)
shuita2333/PAD-codes/MMaDA-PAD/models/logging.py
pointer only (licence: NONE); the stored copy get_code serves, from Feng-Hong/WINO-DLLM, is pointer only (licence: NONE) · get_code("376bf22d5737fc2a")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

get_logits Not yet run
unverified (harvested, no recorded run)
shuita2333/PAD-codes/LLaDA-PAD/get_log_likelihood.py
pointer only (licence: NONE); the stored copy get_code serves, from ml-gsai/llada, is pointer only (licence: NONE) · get_code("bc58c88b365ae743")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, shuita2333/PAD-codes when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

Also present in other papers' code

Of 26 samples lifted from this paper's linked repositories, 26 carry a function identity and 0 do not. Of the 25 distinct functions with an identity, 20 are also present in other papers whose code we hold (20 with at least one paper dated earlier, 5 the same month, 13 later) and 5 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
nonlinearity 156 other papers
151 dated earlier, 1 the same month, 4 later · 25 only as an assembly brick
2104.07652 (2021-04)
Geometry-Free View Synthesis: Transformers and no 3D Priors
2106.00132 (2021-06)
On Fast Sampling of Diffusion Probabilistic Models
2106.06819 (2021-06)
D2C: Diffusion-Denoising Models for Few-shot Conditional Generation
and 153 more not named here
Normalize 90 other papers
86 dated earlier, 1 the same month, 3 later · 32 only as an assembly brick
2104.07652 (2021-04)
Geometry-Free View Synthesis: Transformers and no 3D Priors
2106.00132 (2021-06)
On Fast Sampling of Diffusion Probabilistic Models
2106.06819 (2021-06)
D2C: Diffusion-Denoising Models for Few-shot Conditional Generation
and 87 more not named here
log
brick of this paper's assembly: helper
22 other papers
21 dated earlier, 1 later · 9 only as an assembly brick
1611.07004 (2016-11)
Image-to-Image Translation with Conditional Adversarial Networks
1907.05242 (2019-07)
Large Memory Layers with Product Keys
2203.07852 (2022-03)
Block-Recurrent Transformers
and 19 more not named here
get_linear_schedule_with_warmup 21 other papers
19 dated earlier, 1 the same month, 1 later
2004.10964 (2020-04)
Don't Stop Pretraining: Adapt Language Models to Domains and Tasks
2011.07430 (2020-11)
Audio-Visual Event Recognition through the lens of Adversary
2109.04912 (2021-09)
ReasonBERT: Pre-trained to Reason with Distant Supervision
and 18 more not named here
add_gumbel_noise 18 other papers
11 dated earlier, 1 the same month, 6 later · 3 only as an assembly brick
2502.09992 (2025-02)
Large Language Diffusion Models
2505.15809 (2025-05)
MMaDA: Multimodal Large Diffusion Language Models
2506.06295 (2025-06)
dLLM-Cache: Accelerating Diffusion Large Language Models with Adaptive Caching
and 15 more not named here
causal_attention_bias 16 other papers
10 dated earlier, 2 the same month, 4 later
2406.10670 (2024-06)
CoLoR-Filter: Conditional Loss Reduction Filtering for Targeted Language Model Pre-training
2409.17146 (2024-09)
Molmo and PixMo: Open Weights and Open Data for State-of-the-Art Vision-Language Models
2410.01380 (2024-10)
Knowledge Entropy Decay during Language Model Pretraining Hinders New Knowledge Acquisition
and 13 more not named here
get_num_transfer_tokens 15 other papers
10 dated earlier, 5 later · 1 only as an assembly brick
2502.09992 (2025-02)
Large Language Diffusion Models
2505.15809 (2025-05)
MMaDA: Multimodal Large Diffusion Language Models
2506.06295 (2025-06)
dLLM-Cache: Accelerating Diffusion Large Language Models with Adaptive Caching
and 12 more not named here
get_constant_schedule 13 other papers
12 dated earlier, 1 later
2004.10964 (2020-04)
Don't Stop Pretraining: Adapt Language Models to Domains and Tasks
2204.13902 (2022-04)
Fast Sampling of Diffusion Models with Exponential Integrator
2305.15328 (2023-05)
Visual Programming for Text-to-Image Generation and Evaluation
and 10 more not named here
get_constant_schedule_with_warmup 13 other papers
12 dated earlier, 1 later
2004.10964 (2020-04)
Don't Stop Pretraining: Adapt Language Models to Domains and Tasks
2204.13902 (2022-04)
Fast Sampling of Diffusion Models with Exponential Integrator
2305.15328 (2023-05)
Visual Programming for Text-to-Image Generation and Evaluation
and 10 more not named here
get_logger 12 other papers
11 dated earlier, 1 later
2004.10964 (2020-04)
Don't Stop Pretraining: Adapt Language Models to Domains and Tasks
2305.01938 (2023-05)
Doc2SoarGraph: Discrete Reasoning over Visually-Rich Table-Text Documents via Semantic-Oriented Hierarchical Graphs
2406.05478 (2024-06)
Revisiting Non-Autoregressive Transformers for Efficient Image Synthesis
and 9 more not named here
unpack_time 6 other papers
5 dated earlier, 1 later
2503.03651 (2025-03)
DoraCycle: Domain-Oriented Adaptation of Unified Generative Model in Multimodal Cycles
2505.14671 (2025-05)
UniCTokens: Boosting Personalized Understanding and Generation via Unified Concept Tokens
2505.15809 (2025-05)
MMaDA: Multimodal Large Diffusion Language Models
and 3 more not named here
broadcast 4 other papers
3 dated earlier, 1 later
2503.03651 (2025-03)
DoraCycle: Domain-Oriented Adaptation of Unified Generative Model in Multimodal Cycles
2505.14671 (2025-05)
UniCTokens: Boosting Personalized Understanding and Generation via Unified Concept Tokens
2507.11097 (2025-07)
The Devil behind the mask: An emergent safety vulnerability of Diffusion LLMs
and 1 more not named here
parse_structured 4 other papers
3 dated earlier, 1 later
2503.03651 (2025-03)
DoraCycle: Domain-Oriented Adaptation of Unified Generative Model in Multimodal Cycles
2505.14671 (2025-05)
UniCTokens: Boosting Personalized Understanding and Generation via Unified Concept Tokens
2507.11097 (2025-07)
The Devil behind the mask: An emergent safety vulnerability of Diffusion LLMs
and 1 more not named here
get_parameter_device 3 other papers
3 dated earlier
2505.14671 (2025-05)
UniCTokens: Boosting Personalized Understanding and Generation via Unified Concept Tokens
2506.15564 (2025-06)
Show-o2: Improved Native Unified Multimodal Models
2507.11097 (2025-07)
The Devil behind the mask: An emergent safety vulnerability of Diffusion LLMs
get_parameter_dtype 3 other papers
3 dated earlier
2505.14671 (2025-05)
UniCTokens: Boosting Personalized Understanding and Generation via Unified Concept Tokens
2506.15564 (2025-06)
Show-o2: Improved Native Unified Multimodal Models
2507.11097 (2025-07)
The Devil behind the mask: An emergent safety vulnerability of Diffusion LLMs
forward_process 2 other papers
2 dated earlier
2502.09992 (2025-02)
Large Language Diffusion Models
2506.15735 (2025-06)
ContextBench: Modifying Contexts for Targeted Latent Activation
get_log_likelihood 2 other papers
2 dated earlier
2502.09992 (2025-02)
Large Language Diffusion Models
2506.15735 (2025-06)
ContextBench: Modifying Contexts for Targeted Latent Activation
get_logits 2 other papers
2 dated earlier
2502.09992 (2025-02)
Large Language Diffusion Models
2506.15735 (2025-06)
ContextBench: Modifying Contexts for Targeted Latent Activation
gumbel_noise
brick of this paper's assembly: helper
1 other paper
1 dated earlier
2505.14671 (2025-05)
UniCTokens: Boosting Personalized Understanding and Generation via Unified Concept Tokens
gumbel_sample
brick of this paper's assembly: entry
1 other paper
1 dated earlier
2505.14671 (2025-05)
UniCTokens: Boosting Personalized Understanding and Generation via Unified Concept Tokens

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Diffusion language models (DLMs) have recently emerged as a competitive alternative to autoregressive (AR) models, offering parallel decoding, competitive generation quality, and initial evidence of improved jailbreak robustness. Despite this progress, the role of sampling mechanisms in shaping refusal behavior remains poorly understood. To address this gap, we present a comprehensive study of step-wise refusal dynamics. We show that diffusion remasking can promote recovery from harmful intermediate generations, provide evidence that this behavior is tied to the sampling mechanism, and demonstrate that switching from AR to diffusion sampling improves jailbreak robustness, including under fixed model weights. To capture generation dynamics not observable at the text level, we propose the Step-Wise Refusal Internal Dynamics (SRI) signal. Consistent with our text-level findings, SRI shows that recovery fails primarily under AR sampling, with these failures often appearing anomalous relative to harmless generations in the SRI space. Based on this observation, we show that SRI enables a simple jailbreak detector that does not modify inference and generalizes to unseen attacks by training only on benign SRI signals. Our evaluation shows that this detector matches or outperforms existing jailbreak detection baselines while adding negligible overhead.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2602.02600")
get_code_for_paper("2602.02600")
have("2602.02600")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2602.02600.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2602.02600.svg)](https://syntology.ai/paper/2602.02600)

Connect an agent — have() is free.