SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2605.06407 · 2026

WavCube: Unifying Speech Representation for Understanding and Generation via Semantic-Acoustic Joint Modeling

Xie Chen, Tian Tan, Qian Chen, Shan Yang, Yifan Yang, Yakun Song, Ziyang Ma, Wenxi Chen, Zhikang Niu, Guanrou Yang, Wenrui Liu, Yushen Chen, and 2 more

arXiv · PDF · Open in the Atlas

Code that ran

7 of 8 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 0 have no contract verdict; 7 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/2605.06407.json · badge: /api/ran/2605.06407.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
yanghaha0908/WavCube canonical (classifier label)
classifier basis: census
7 ran · 1 no recorded run
FunctionStatusWhere it lives
characterize Ran
ran; our contract draft was wrong, not the code
yanghaha0908/WavCube/evaluation/compute_wer.py
code served (permissive licence) · get_code("ad9a15f964a5b45f")
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 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 re
import sys
import unicodedata

spacelist = [' ', '\t', '\r', '\n']

puncts = ['!', ',', '?', '、', '。', '!', ',', ';', '?', ':', '「', '」', '︰', '『', '』', '《', '》']

def characterize(string):
    res = []
    i = 0
    while i < len(string):
        char = string[i]
        if char in puncts:
            i += 1
            continue
        cat1 = unicodedata.category(char)
        if cat1 == 'Zs' or cat1 == 'Cn' or char in spacelist:
            i += 1
            continue
        if cat1 == 'Lo':
            res.append(char)
            i += 1
        else:
            sep = ' '
            if char == '<':
                sep = '>'
            j = i + 1
            while j < len(string):
                c = string[j]
                if ord(c) >= 128 or c in spacelist or c == sep:
                    break
                j += 1
            if j < len(string) and string[j] == '>':
                j += 1
            res.append(string[i:j])
            i = j
    return res

# Test cases for the function
test_string_1 = "Hello, world! This is a test."
test_string_2 = "test<tag>data 中文 text"
test_string_3 = "abc123def!ghi?jkl"
test_string_4 = ""

ARGS = {
    "string": "Hello world test<data> 中文"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
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 list, ndim 1, dtype family float — 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
load_feature Ran
ran; our contract draft was wrong, not the code
yanghaha0908/WavCube/feature_to_wav.py
code served (permissive licence) · get_code("10ae85266c7ccbf9")
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
from pathlib import Path
import tempfile
import os

torch.manual_seed(0)

# Create a temporary .pt file with a valid payload
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, "test_feature.pt")

# Create a feature tensor [T_feat=5, 128]
feature = torch.randn(5, 128)
payload = {
    'feature': feature,
    'sample_rate': 16000,
    'source': 'test_feature.pt'
}
torch.save(payload, temp_file)

ARGS = {
    'path': temp_file,
    'device': 'cpu'
}
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 — 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
normalize Ran
ran; our contract draft was wrong, not the code
yanghaha0908/WavCube/evaluation/compute_wer.py
code served (permissive licence) · get_code("e14967660d98560c")
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 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 numpy as np

torch.manual_seed(0)

# Analyze the function:
# - sentence: iterated over (needs to be iterable of tokens/strings)
# - ignore_words: used with 'in' operator (needs to be a set or dict-like)
# - cs: used in boolean context (needs to be bool or truthy/falsy)
# - split: used with 'in' operator and indexing (needs to be dict or None)

# Construct valid arguments
sentence = ["HELLO", "world", "<tag>text</tag>", "ignore_me", "split_word"]
ignore_words = {"ignore_me"}
cs = False  # case sensitive flag
split = {"split_word": ["split", "into", "parts"]}

ARGS = {
    "sentence": sentence,
    "ignore_words": ignore_words,
    "cs": cs,
    "split": split
}
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 float — 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
safe_log Ran
ran; our contract draft was wrong, not the code
yanghaha0908/WavCube/vocos/modules.py
code served (permissive licence) · get_code("5923f41e711c0b69")
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 with values that need clipping
x = torch.tensor([[0.5, 1e-8, 0.1, 2.0],
                   [0.3, 0.7, 1e-9, 0.05]], dtype=torch.float32)

# Define arguments dictionary
ARGS = {
    "x": x,
    "clip_val": 1e-07
}
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
stripoff_tags Ran
ran; our contract draft was wrong, not the code
yanghaha0908/WavCube/evaluation/compute_wer.py
code served (permissive licence) · get_code("32ade8e8e8927d3e")
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 str. 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

def stripoff_tags(x):
    if not x:
        return ''
    chars = []
    i = 0
    T = len(x)
    while i < T:
        if x[i] == '<':
            while i < T and x[i] != '>':
                i += 1
            i += 1
        else:
            chars.append(x[i])
            i += 1
    return ''.join(chars)

ARGS = {
    'x': 'hello <tag>world</tag> foo'
}
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 str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
symexp Ran
ran; our contract draft was wrong, not the code
yanghaha0908/WavCube/vocos/modules.py
code served (permissive licence) · get_code("f1abe87534e9b4b5")
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)

# The function symexp takes a single tensor argument and applies:
# torch.sign(x) * (torch.exp(x.abs()) - 1)
# It needs a tensor input

x = torch.randn(2, 4)

ARGS = {
    "x": x,
}
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
symlog Ran
ran; our contract draft was wrong, not the code
yanghaha0908/WavCube/vocos/modules.py
code served (permissive licence) · get_code("3acb919a94ac1a19")
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)

# The symlog function takes a single tensor argument and applies
# element-wise symmetric logarithm: sign(x) * log1p(|x|)
# We need to provide a small tensor to test this.

x = torch.randn(2, 4)

ARGS = {
    "x": x,
}
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
ECAPA_TDNN_SMALL Not yet run
unverified (harvested, no recorded run)
yanghaha0908/WavCube/evaluation/ecapa_tdnn.py
code served (permissive licence) · get_code("bea543d0e76dfc84")
How it ran

no invocation recorded.

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 distinct functions with an identity, 4 are also present in other papers whose code we hold (4 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.

FunctionAlso present inEarliest-datedOthers named
safe_log 5 other papers
5 dated earlier
2306.00814 (2023-06)
Vocos: Closing the gap between time-domain and Fourier-based neural vocoders for high-quality audio synthesis
2402.12208 (2024-02)
Language-Codec: Bridging Discrete Codec Representations and Speech Language Models
2405.11273 (2024-05)
Uni-MoE: Scaling Unified Multimodal LLMs with Mixture of Experts
and 2 more not named here
symexp 5 other papers
5 dated earlier
2306.00814 (2023-06)
Vocos: Closing the gap between time-domain and Fourier-based neural vocoders for high-quality audio synthesis
2402.12208 (2024-02)
Language-Codec: Bridging Discrete Codec Representations and Speech Language Models
2405.11273 (2024-05)
Uni-MoE: Scaling Unified Multimodal LLMs with Mixture of Experts
and 2 more not named here
symlog 5 other papers
5 dated earlier
2306.00814 (2023-06)
Vocos: Closing the gap between time-domain and Fourier-based neural vocoders for high-quality audio synthesis
2402.12208 (2024-02)
Language-Codec: Bridging Discrete Codec Representations and Speech Language Models
2405.11273 (2024-05)
Uni-MoE: Scaling Unified Multimodal LLMs with Mixture of Experts
and 2 more not named here
ECAPA_TDNN_SMALL 1 other paper
1 dated earlier
2410.06885 (2024-10)
F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching

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

Integrating speech understanding and generation is a pivotal step toward building unified speech models. However, the different representations required for these two tasks currently pose significant compatibility challenges. Typically, semantics-oriented features are learned from self-supervised learning (SSL), and acoustic-oriented features from reconstruction. Such fragmented representations hinder the realization of truly unified speech systems. We present WavCube, a compact continuous latent derived from an SSL speech encoder that simultaneously supports speech understanding, reconstruction, and generation. WavCube employs a two-stage training scheme. Stage 1 trains a semantic bottleneck to filter off-manifold redundancy that makes raw SSL features intractable for diffusion. Stage 2 injects fine-grained acoustic details via end-to-end reconstruction, while a semantic anchoring loss ensures the representation remains grounded within its original semantic manifold. Comprehensive experiments show that WavCube closely approaches WavLM performance on SUPERB despite an 8× dimensional compression, attains reconstruction quality on par with existing acoustic representations, delivers state-of-the-art zero-shot TTS performance with markedly faster training convergence, and excels in speech enhancement, separation, and voice conversion tasks on the SUPERB-SG benchmark. Systematic ablations reveal that WavCube's two-stage recipe resolves two intrinsic flaws of SSL features for generative modeling, paving the way for future unified speech systems. Codes and checkpoints are available at https://github.com/yanghaha0908/WavCube.

For agents

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

get_harvested_code_for_paper("2605.06407")
get_code_for_paper("2605.06407")
have("2605.06407")

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

curl https://syntology.ai/api/ran/2605.06407.json

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

[![Syntology run record](https://syntology.ai/api/ran/2605.06407.svg)](https://syntology.ai/paper/2605.06407)

Connect an agent — have() is free.