SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2606.01062 · ICML · 2026

DAG-MoE: From Simple Mixture to Structural Aggregation in Mixture-of-Experts

Yixin Chen, Yinglong Xia, Hanqing Zeng, Ren Chen, Qiang Zhang, Jiarui Feng, Karish Grover, Ruizhong Qiu, Qifan Wang, Zhuokai Zhao, Dongqi Fu, Jiayi Liu, and 2 more

arXiv · PDF · Open in the Atlas

Code that ran

1 of 4 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, 0 violated it and 1 has no contract verdict; 0 failures of our instrument; 3 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/2606.01062.json · badge: /api/ran/2606.01062.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
JiaruiFeng/DAG-MoE community
Papers with Code archive
1 ran · 3 no recorded run
FunctionStatusWhere it lives
DAGMoEMLP Ran
ran on a synthesized input
JiaruiFeng/DAG-MoE/model/dag_moe/modeling_dagmoe.py
code served (permissive licence) · get_code("dd5eb2f515700590")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 3, dtype family float, shape (2, 3, 6) 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.

from typing import Callable, Optional, Tuple, Union, List
import torch
from torch import nn
from transformers.activations import ACT2FN
from transformers.models.llama import LlamaConfig

class DAGMoEConfig(LlamaConfig):
    model_type = 'dag_moe'

    def __init__(self, hidden_size=4096, intermediate_size=11008, num_hidden_layers=32, num_attention_heads=32, num_key_value_heads=None, head_dim=None, **kwargs):
        super().__init__(hidden_size=hidden_size, intermediate_size=intermediate_size, num_hidden_layers=num_hidden_layers, num_attention_heads=num_attention_heads, num_key_value_heads=num_key_value_heads, head_dim=head_dim, **kwargs)

class DAGMoEMLP(nn.Module):

    def __init__(self, config: DAGMoEConfig, hidden_size: int, intermediate_size: int, output_size: Optional[int]=None):
        super().__init__()
        self.hidden_size = hidden_size
        self.intermediate_size = intermediate_size
        self.output_size = output_size if output_size is not None else hidden_size
        self.gate_proj = nn.Linear(hidden_size, intermediate_size, bias=config.mlp_bias)
        self.up_proj = nn.Linear(hidden_size, intermediate_size, bias=config.mlp_bias)
        self.down_proj = nn.Linear(intermediate_size, self.output_size, bias=config.mlp_bias)
        self.act_fn = ACT2FN[config.hidden_act]

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
        return down_proj


config = DAGMoEConfig(
    hidden_size=4,
    intermediate_size=8,
    num_hidden_layers=2,
    num_attention_heads=2,
    mlp_bias=True,
    hidden_act='silu'
)

INIT_ARGS = {
    'config': config,
    'hidden_size': 4,
    'intermediate_size': 8,
    'output_size': 6
}

ARGS = {
    'x': torch.randn(2, 3, 4)
}
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
Imports
torch, transformers — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 3, dtype family float, shape (2, 3, 6) — shape (2, 3, 6) 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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
DAGMoEBlock Not yet run
unverified (harvested, no recorded run)
JiaruiFeng/DAG-MoE/model/dag_moe/modeling_dagmoe.py
code served (permissive licence) · get_code("b3941851a0e470b5")
How it ran

no invocation recorded.

DAGMoEConfig Not yet run
unverified (harvested, no recorded run)
JiaruiFeng/DAG-MoE/model/dag_moe/modeling_dagmoe.py
code served (permissive licence) · get_code("28a4299982e5baa6")
How it ran

no invocation recorded.

DAGMoERouter Not yet run
unverified (harvested, no recorded run)
JiaruiFeng/DAG-MoE/model/dag_moe/modeling_dagmoe.py
code served (permissive licence) · get_code("94201f1401e96ee4")
How it ran

no invocation recorded.

Also present in other papers' code

Of 4 samples lifted from this paper's linked repositories, 4 carry a function identity and 0 do not. Of the 4 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 4 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.

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

Mixture-of-Experts (MoE) models have become a leading approach for decoupling parameter count from computational cost in large language models, yet effectively scaling MoE performance remains a challenge. Prior work shows that fine-grained experts enlarge the space of expert combinations and improve flexibility, but they also impose substantial routing overhead, creating a new scalability bottleneck. In this paper, we explore a complementary axis for scaling-how expert outputs are aggregated. We theoretically show that replacing the standard weighted-summation aggregation with structural aggregation expands the expertcombination space without altering the experts or router, and enables possible multi-step reasoning within a single MoE layer. To this end, we propose DAG-MoE, a sparse MoE framework that employs a lightweight module to automatically learn the optimal aggregation structure among the selected experts. Extensive experiments under standard language modeling settings show that DAG-MoE consistently improves performance in both pretraining and fine-tuning, surpassing traditional MoE baselines.

For agents

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

get_harvested_code_for_paper("2606.01062")
get_code_for_paper("2606.01062")
have("2606.01062")

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

curl https://syntology.ai/api/ran/2606.01062.json

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

[![Syntology run record](https://syntology.ai/api/ran/2606.01062.svg)](https://syntology.ai/paper/2606.01062)

Connect an agent — have() is free.