SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2602.06953 · 2026

DAWN: Dependency-Aware Fast Inference for Diffusion LLMs

Tianwei Zhang, Zhi Wang, Wenya Wang, Jiajun Luo, Shen Ren, Lizhuo Luo, Zhuoran Shi

arXiv · PDF · Open in the Atlas

Code that ran

16 of 17 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; 0 honoured the contract we drafted, 1 violated it and 0 have no contract verdict; 15 failures of our instrument; 1 no recorded run), 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/2602.06953.json · badge: /api/ran/2602.06953.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
lizhuo-luo/DAWN canonical (classifier label)
classifier basis: census
16 ran · 1 no recorded run
FunctionStatusWhere it lives
apply_rotary_pos_emb Ran
ran; our contract draft was wrong, not the code
lizhuo-luo/DAWN/dream/model/modeling_dream.py
pointer only (licence: MIT); the stored copy get_code serves, from GATECH-EIC/R2-dLLM, is pointer only (licence: NONE) · get_code("ac6507e437413e71")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from GATECH-EIC/R2-dLLM, and GitHub detected no licence in that repository 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.

attn_avg Ran
ran; our contract draft was wrong, not the code
lizhuo-luo/DAWN/dream/model/modeling_dream.py
code served (permissive licence) · get_code("dfc107aa05e78b6d")
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, 3) 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
import math

torch.manual_seed(0)

# q: query tensor, shape (batch, seq_len, d_model)
# k: key tensor, shape (batch, seq_len, d_model)
# attn_mask: optional attention mask, same shape as attention scores

batch_size = 2
seq_len = 3
d_model = 4

q = torch.randn(batch_size, seq_len, d_model)
k = torch.randn(batch_size, seq_len, d_model)

# attn_mask should have shape compatible with (batch, seq_len, seq_len)
# It gets added to the attention scores, so it needs the right shape
attn_mask = torch.zeros(batch_size, seq_len, seq_len)

ARGS = {
    "q": q,
    "k": k,
    "attn_mask": attn_mask,
}
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, 3) — shape (2, 3) 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
build_state Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
lizhuo-luo/DAWN/demo/viz.py
code served (permissive licence) · get_code("b8100a269ea0a0fc")
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 tuple. The image it ran in was not recorded. If your copy returns the same type 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
import numpy as np

torch.manual_seed(0)

# Define a decode_token function that converts token IDs to strings
def decode_token(tid):
    """Convert token ID to string representation."""
    token_map = {
        0: "hello",
        1: "world",
        2: "test",
        3: "[MASK]",
        4: "</s>",
        5: "token",
    }
    return token_map.get(tid, f"token_{tid}")

# Create test arguments
now_ids = torch.tensor([0, 1, 2, 3, 4, 5], dtype=torch.long)
mask_id = 3
eos_id = 4

ARGS = {
    "now_ids": now_ids,
    "decode_token": decode_token,
    "mask_id": mask_id,
    "eos_id": eos_id,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
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
detect_attn_sinks Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
lizhuo-luo/DAWN/dream/model/gdllm_utils.py
code served (permissive licence) · get_code("bfd66364f5c97c6c")
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 attention scores tensor with shape (B, L, L)
# B=2 (batch size), L=4 (sequence length)
attn_scores = torch.randn(2, 4, 4)

ARGS = {
    "attn_scores": attn_scores,
    "ratio": 0.5,
    "topk": 2
}
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
detect_attn_sinks_ Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
lizhuo-luo/DAWN/dream/model/gdllm_utils.py
code served (permissive licence) · get_code("b53232e0e49f5901")
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
import torch.nn as nn

torch.manual_seed(0)

# Create attention scores tensor
# Shape: [B, S, S] where B=batch_size, S=sequence_length
# attn_scores should be softmax attention (values in [0, 1])
B = 2
S = 4
attn_scores = torch.rand(B, S, S)
attn_scores = torch.softmax(attn_scores, dim=-1)

threshold = 0.02

ARGS = {
    "attn_scores": attn_scores,
    "threshold": threshold,
}
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
extract_longest_valid_code Ran
ran; our contract draft was wrong, not the code
lizhuo-luo/DAWN/dream/sanitize.py
pointer only (licence: MIT); the stored copy get_code serves, from GATECH-EIC/R2-dLLM, is pointer only (licence: NONE) · get_code("044f03d955ca1151")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from GATECH-EIC/R2-dLLM, and GitHub detected no licence in that repository 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.

frames_from_snapshots Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
lizhuo-luo/DAWN/demo/viz.py
code served (permissive licence) · get_code("316093b5a11a1386")
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 list. The image it ran in was not recorded. If your copy returns the same type 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)

# Define a simple decode_token function that converts token IDs to strings
def decode_token(tid):
    """Convert token ID to string representation."""
    token_map = {
        0: "hello",
        1: "world",
        2: "foo",
        3: "bar",
        4: "baz",
        5: "[MASK]",
    }
    return token_map.get(tid, f"token_{tid}")

# Create sample snapshots: list of (1, L) tensors
# prompt_len = 2, gen_len = 3, so we need tensors of shape (1, 5)
# Each tensor represents a snapshot at a generation step
xs = [
    torch.tensor([[5, 5, 5, 5, 5]], dtype=torch.long),  # Initial: all masked
    torch.tensor([[5, 5, 0, 5, 5]], dtype=torch.long),  # Step 1: first token decoded
    torch.tensor([[5, 5, 0, 1, 5]], dtype=torch.long),  # Step 2: second token decoded
    torch.tensor([[5, 5, 0, 1, 2]], dtype=torch.long),  # Step 3: third token decoded
]

prompt_len = 2
gen_len = 3
mask_id = 5
eos_id = None

ARGS = {
    "xs": xs,
    "prompt_len": prompt_len,
    "gen_len": gen_len,
    "decode_token": decode_token,
    "mask_id": mask_id,
    "eos_id": eos_id,
}
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 list — a list output has no array shape of its own; none is stated
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
get_pairs Ran
ran; our contract draft was wrong, not the code
lizhuo-luo/DAWN/dream/model/tokenization_dream.py
code served (permissive licence) · get_code("c083e60f5001bd38")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned set. The image it ran in was not recorded. If your copy returns the same type 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.

# The function expects a word represented as a tuple of symbols (strings)
# It iterates through the word starting from index 0 and 1, creating pairs of consecutive symbols

word = ("h", "e", "ll", "o")

ARGS = {
    "word": word
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type set — a set output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a set output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
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
make_decode_token Ran
ran; our contract draft was wrong, not the code
lizhuo-luo/DAWN/demo/viz.py
code served (permissive licence) · get_code("10811457e1fd0935")
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 function. The image it ran in was not recorded. If your copy returns the same type 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
import torch.nn as nn

# Create a minimal tokenizer stand-in class
class MinimalTokenizer:
    def __init__(self):
        self.vocab = {
            0: "hello",
            1: "world",
            2: "",  # Empty token to test the fallback to '·'
            3: "test"
        }
    
    def decode(self, token_ids, skip_special_tokens=True):
        """Decode a list of token IDs to a string."""
        result = []
        for tid in token_ids:
            if tid in self.vocab:
                token_str = self.vocab[tid]
                if skip_special_tokens and token_str == "":
                    continue
                result.append(token_str)
        decoded = " ".join(result) if result else ""
        return decoded

tokenizer = MinimalTokenizer()

ARGS = {
    "tokenizer": tokenizer
}
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 function — a function output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a function output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
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
refine_text Ran
ran; our contract draft was wrong, not the code
lizhuo-luo/DAWN/dream/sanitize.py
pointer only (licence: MIT); the stored copy get_code serves, from lux0926/asprm, is pointer only (licence: NONE) · get_code("a1a9545f89348999")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from lux0926/asprm, and GitHub detected no licence in that repository 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.

rotate_half Ran
ran; our contract draft was wrong, not the code
lizhuo-luo/DAWN/dream/model/modeling_dream.py
code served (permissive licence) · get_code("b99eea6376d1e212")
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 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)

# The function rotate_half takes a single parameter 'x' which is a tensor
# It accesses x.shape[-1] to get the last dimension
# It uses indexing with [..., :] slicing notation
# It calls torch.cat() to concatenate tensors

# Create a small tensor with shape (2, 4) - batch size 2, hidden dim 4
x = torch.randn(2, 4)

ARGS = {
    "x": x
}
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 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
sample_tokens Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
lizhuo-luo/DAWN/dream/model/generation_utils.py
code served (permissive licence) · get_code("1891fed3a1008763")
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 tuple, ndim 2, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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
import torch.nn.functional as F

torch.manual_seed(0)

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

ARGS = {
    "logits": logits,
    "temperature": 0.8,
    "top_p": 0.9,
    "top_k": 3,
    "margin_confidence": False,
    "neg_entropy": False,
}
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 tuple, ndim 2, dtype family float — a tuple output has no array shape of its own; none is stated
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
select_parallel_tokens_conflict_mis Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
lizhuo-luo/DAWN/dream/model/gdllm_utils.py
code served (permissive licence) · get_code("cc33a03c053c97c6")
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 list, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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
import numpy as np

torch.manual_seed(0)

# Construct tiny tensors for the function
# edge_mask: adjacency matrix representing conflicts between nodes
# Should be a square boolean tensor
num_nodes = 4
edge_mask = torch.zeros((num_nodes, num_nodes), dtype=torch.bool)
# Add some sample conflicts
edge_mask[0, 1] = True
edge_mask[1, 2] = True
edge_mask[2, 3] = True

# node_mask: boolean mask indicating which nodes are active
node_mask = torch.ones(num_nodes, dtype=torch.bool)

# confidence: scores for each node to prioritize selection
confidence = torch.tensor([0.9, 0.7, 0.5, 0.3], dtype=torch.float32)

# max_parallel: maximum number of parallel tokens to select
max_parallel = 2

ARGS = {
    "edge_mask": edge_mask,
    "node_mask": node_mask,
    "confidence": confidence,
    "max_parallel": max_parallel,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family int — a list output has no array shape of its own; none is stated
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
syntax_check Ran
ran, violated the contract we drafted
lizhuo-luo/DAWN/dream/sanitize.py
pointer only (licence: MIT); the stored copy get_code serves, from GATECH-EIC/R2-dLLM, is pointer only (licence: NONE) · get_code("d7d30a99e0029e36")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from GATECH-EIC/R2-dLLM, and GitHub detected no licence in that repository 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.

top_k_logits Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
lizhuo-luo/DAWN/dream/model/generation_utils.py
pointer only (licence: MIT); the stored copy get_code serves, from xiaoshideta/Streaming-dLLM, is pointer only (licence: NONE) · get_code("d78ba22cffe4710b")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from xiaoshideta/Streaming-dLLM, and GitHub detected no licence in that repository 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.

top_p_logits Ran
ran; our contract draft was wrong, not the code
lizhuo-luo/DAWN/dream/model/generation_utils.py
pointer only (licence: MIT); the stored copy get_code serves, from xiaoshideta/Streaming-dLLM, is pointer only (licence: NONE) · get_code("8139ba2bcf3eb84b")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from xiaoshideta/Streaming-dLLM, and GitHub detected no licence in that repository 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_backend Not yet run
unverified (harvested, no recorded run)
lizhuo-luo/DAWN/demo/registry.py
code served (permissive licence) · get_code("1e9e8d1e3c7a52db")
How it ran

no invocation recorded.

Also present in other papers' code

Of 17 samples lifted from this paper's linked repositories, 17 carry a function identity and 0 do not. Of the 17 distinct functions with an identity, 9 are also present in other papers whose code we hold (7 with at least one paper dated earlier, 5 the same month, 9 later) and 8 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
rotate_half 360 other papers
324 dated earlier, 6 the same month, 27 later, 3 undated · 30 only as an assembly brick
1901.02860 (2019-01)
Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context
2203.08913 (2022-03)
Memorizing Transformers
2302.13971 (2023-02)
LLaMA: Open and Efficient Foundation Language Models
and 357 more not named here
get_pairs 258 other papers
242 dated earlier, 2 the same month, 13 later, 1 undated · 4 only as an assembly brick
1802.06474 (2018-02)
A Closed-form Solution to Photorealistic Image Stylization
1803.01128 (2018-03)
Seq2Sick: Evaluating the Robustness of Sequence-to-Sequence Models with Adversarial Examples
1809.00125 (2018-09)
Simple Fusion: Return of the Language Model
and 255 more not named here
top_k_logits 10 other papers
3 dated earlier, 2 the same month, 4 later, 1 undated · 5 only as an assembly brick
2202.04200 (2022-02)
MaskGIT: Masked Generative Image Transformer
2509.26432 (2025-09)
AdaBlock-dLLM: SEMANTIC-AWARE DIFFUSION LLM INFERENCE VIA ADAPTIVE BLOCK SIZE
2601.17917 (2026-01)
Streaming-dLLM: Accelerating Diffusion LLMs via Suffix Pruning and Dynamic Decoding
and 7 more not named here
top_p_logits 10 other papers
3 dated earlier, 2 the same month, 4 later, 1 undated · 5 only as an assembly brick
2202.04200 (2022-02)
MaskGIT: Masked Generative Image Transformer
2509.26432 (2025-09)
AdaBlock-dLLM: SEMANTIC-AWARE DIFFUSION LLM INFERENCE VIA ADAPTIVE BLOCK SIZE
2601.17917 (2026-01)
Streaming-dLLM: Accelerating Diffusion LLMs via Suffix Pruning and Dynamic Decoding
and 7 more not named here
sample_tokens 5 other papers
1 dated earlier, 1 the same month, 3 later · 1 only as an assembly brick
2509.26432 (2025-09)
AdaBlock-dLLM: SEMANTIC-AWARE DIFFUSION LLM INFERENCE VIA ADAPTIVE BLOCK SIZE
2602.01326 (2026-02)
DreamOn: Diffusion Language Models For Code Infilling Beyond Fixed-size Canvas
2604.15750 (2026-04)
DepCap: Adaptive Block-Wise Parallel Decoding for Efficient Diffusion LM Inference
and 2 more not named here
syntax_check 5 other papers
3 dated earlier, 2 later
2406.15877 (2024-06)
BigCodeBench: Benchmarking Code Generation with Diverse Function Calls and Complex Instructions
2412.03578 (2024-12)
PerfCodeGen: Improving Performance of LLM Generated Code with Execution Feedback
2502.02827 (2025-02)
COFFE: A Code Efficiency Benchmark for Code Generation
and 2 more not named here
refine_text 3 other papers
1 dated earlier, 2 later
2502.13943 (2025-02)
AdaptiveStep: Automatically Dividing Reasoning Step through Model Confidence
2604.15750 (2026-04)
DepCap: Adaptive Block-Wise Parallel Decoding for Efficient Diffusion LM Inference
2604.18995 (2026-04)
R 2 -dLLM: Accelerating Diffusion Large Language Models via Spatio-Temporal Redundancy Reduction
apply_rotary_pos_emb 2 other papers
2 later
2604.15750 (2026-04)
DepCap: Adaptive Block-Wise Parallel Decoding for Efficient Diffusion LM Inference
2604.18995 (2026-04)
R 2 -dLLM: Accelerating Diffusion Large Language Models via Spatio-Temporal Redundancy Reduction
extract_longest_valid_code 2 other papers
2 later
2604.15750 (2026-04)
DepCap: Adaptive Block-Wise Parallel Decoding for Efficient Diffusion LM Inference
2604.18995 (2026-04)
R 2 -dLLM: Accelerating Diffusion Large Language Models via Spatio-Temporal Redundancy Reduction

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 large language models (dLLMs) have shown advantages in text generation, particularly due to their inherent ability for parallel decoding. However, constrained by the quality-speed trade-off, existing inference solutions adopt conservative parallel strategies, leaving substantial efficiency potential underexplored. A core challenge is that parallel decoding assumes each position can be filled independently, but tokens are often semantically coupled. Thus, the correct choice at one position constrains valid choices at others. Without modeling these inter-token dependencies, parallel strategies produce deteriorated outputs. Motivated by this insight, we propose DAWN, a training-free, dependency-aware decoding method for fast dLLM inference. DAWN extracts token dependencies and leverages two key motivations: (1) positions dependent on unmasked certain positions become more reliable, (2) simultaneously unmasking strongly coupled uncertain positions induces errors. Given those findings, DAWN leverages a dependency graph to select more reliable unmasking positions at each iteration, achieving high parallelism with negligible loss in generation quality. Extensive experiments across multiple models and datasets demonstrate that DAWN speedups the inference by 1.80 -8.06× over baselines while preserving the generation quality. Code is released at https://github.com/lizhuo-luo/DAWN.

For agents

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

get_harvested_code_for_paper("2602.06953")
get_code_for_paper("2602.06953")
have("2602.06953")

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

curl https://syntology.ai/api/ran/2602.06953.json

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

[![Syntology run record](https://syntology.ai/api/ran/2602.06953.svg)](https://syntology.ai/paper/2602.06953)

Connect an agent — have() is free.