SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2402.05706 · NeurIPS · 2024

Paralinguistics-Aware Speech-Empowered Large Language Models for Natural Conversation

Jung-Woo Ha, Sungroh Yoon, Heeseung Kim, Min Kang, Yoo, Soonshin Seo, Kyeongseok Jeong, Ohsung Kwon, Soyoon Kim, Jungwhan Kim, Jaehong Lee, Eunwoo Song, and 1 more

arXiv · PDF · Open in the Atlas

Code that ran

10 of 15 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (5 of them constructed an object rather than computing a result; 1 honoured the contract we drafted, 0 violated it and 6 have no contract verdict; 3 failures of our instrument; 5 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; 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/2402.05706.json · badge: /api/ran/2402.05706.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
naver-ai/usdm canonical (classifier label)
classifier basis: classifier_2025
10 ran · 5 no recorded run
FunctionStatusWhere it lives
Attention Ran
ran on a synthesized input
naver-ai/usdm/src/decoder/voicebox/model/voicebox.py
code served (permissive licence) · get_code("cd64f7d57cb575c2")
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; 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 as nn

class Attention(nn.Module):
    """Multi-headed attention from 'Attention Is All You Need' paper"""

    def __init__(self, embed_dim: int, num_heads: int, dropout: float=0.0, is_decoder: bool=False, bias: bool=True, is_causal: bool=False):
        super().__init__()
        self.embed_dim = embed_dim
        self.num_heads = num_heads
        self.dropout = dropout
        self.head_dim = embed_dim // num_heads
        self.scaling = self.head_dim ** (-0.5)
        self.is_decoder = is_decoder
        self.is_causal = is_causal
        self.k_proj = nn.Linear(embed_dim, embed_dim, bias=bias)
        self.v_proj = nn.Linear(embed_dim, embed_dim, bias=bias)
        self.q_proj = nn.Linear(embed_dim, embed_dim, bias=bias)
        self.out_proj = nn.Linear(embed_dim, embed_dim, bias=bias)

    def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
        return tensor.view(bsz, seq_len, self.num_heads, self.head_dim).transpose(1, 2).contiguous()

    def forward(self, hidden_states, attention_mask=None, alibi_bias=None):
        """Input shape: Batch x Time x Channel"""
        bsz, tgt_len, _ = hidden_states.size()
        query_states = self.q_proj(hidden_states) * self.scaling
        key_states = self._shape(self.k_proj(hidden_states), -1, bsz)
        value_states = self._shape(self.v_proj(hidden_states), -1, bsz)
        proj_shape = (bsz * self.num_heads, -1, self.head_dim)
        query_states = self._shape(query_states, tgt_len, bsz).view(*proj_shape)
        key_states = key_states.reshape(*proj_shape)
        value_states = value_states.reshape(*proj_shape)
        src_len = key_states.size(1)
        attn_weights = torch.bmm(query_states, key_states.transpose(1, 2))
        if attn_weights.size() != (bsz * self.num_heads, tgt_len, src_len):
            raise ValueError(f'Attention weights should be of size {(bsz * self.num_heads, tgt_len, src_len)}, but is {attn_weights.size()}')
        if attention_mask is not None:
            if attention_mask.size() != (bsz, 1, tgt_len, src_len):
                raise ValueError(f'Attention mask should be of size {(bsz, 1, tgt_len, src_len)}, but is {attention_mask.size()}')
            attn_weights = attn_weights.view(bsz, self.num_heads, tgt_len, src_len) + alibi_bias + attention_mask
            attn_weights = attn_weights.view(bsz * self.num_heads, tgt_len, src_len)
        attn_weights = nn.functional.softmax(attn_weights, dim=-1)
        attn_probs = nn.functional.dropout(attn_weights, p=self.dropout, training=self.training)
        attn_output = torch.bmm(attn_probs, value_states)
        if attn_output.size() != (bsz * self.num_heads, tgt_len, self.head_dim):
            raise ValueError(f'`attn_output` should be of size {(bsz * self.num_heads, tgt_len, self.head_dim)}, but is {attn_output.size()}')
        attn_output = attn_output.view(bsz, self.num_heads, tgt_len, self.head_dim)
        attn_output = attn_output.transpose(1, 2)
        attn_output = attn_output.reshape(bsz, tgt_len, self.embed_dim)
        attn_output = self.out_proj(attn_output)
        return attn_output


INIT_ARGS = {
    'embed_dim': 4,
    'num_heads': 2,
    'dropout': 0.1,
    'is_decoder': False,
    'bias': True,
    'is_causal': False,
}

ARGS = {
    'hidden_states': torch.randn(2, 3, 4),
    'attention_mask': torch.zeros(2, 1, 3, 3),
    'alibi_bias': torch.zeros(2, 2, 3, 3),
}

if __name__ == '__main__':
    attention = Attention(**INIT_ARGS)
    output = attention(**ARGS)
    print(f"Output shape: {output.shape}")
    print(f"Expected shape: (2, 3, 4)")
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 3, dtype family float — no output shape is stated: observe_signature.py observed this class with its constructor sizes canonicalised where it could, which the node does not record, so its shape is not stated as this setup block's
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)
FeedForward Ran
ran on a synthesized input · constructed an object; no result was computed
naver-ai/usdm/src/decoder/voicebox/model/voicebox.py
code served (permissive licence) · get_code("9e31ddf294975779")
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 an instance of its own class, FeedForward: the run constructed the object and did not call it, so no result was computed. 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

torch.manual_seed(0)

ARGS = {
    "activation_dropout": 0.1,
    "hidden_size": 4,
    "intermediate_size": 8,
    "hidden_dropout": 0.1,
}
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 FeedForward — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a FeedForward 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
GELUActivation Ran
ran on a synthesized input · constructed an object; no result was computed
naver-ai/usdm/src/decoder/voicebox/model/voicebox.py
code served (permissive licence) · get_code("76ec3a4a379092d1")
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 an instance of its own class, GELUActivation: the run constructed the object and did not call it, so no result was computed. 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 math
import torch.nn as nn
from torch import Tensor

torch.manual_seed(0)

ARGS = {
    "use_gelu_python": 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 GELUActivation — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a GELUActivation 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
PositionalConvEmbedding Ran
ran on a synthesized input · constructed an object; no result was computed
naver-ai/usdm/src/decoder/voicebox/model/voicebox.py
code served (permissive licence) · get_code("4ef3da2aa5df518b")
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 an instance of its own class, PositionalConvEmbedding: the run constructed the object and did not call it, so no result was computed. 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
import math

torch.manual_seed(0)

# Parameters for PositionalConvEmbedding.__init__
hidden_size = 4
convpos_width = 3
convpos_groups = 2

ARGS = {
    "hidden_size": hidden_size,
    "convpos_width": convpos_width,
    "convpos_groups": convpos_groups,
}
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 PositionalConvEmbedding — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a PositionalConvEmbedding 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
SamePadLayer Ran
ran on a synthesized input · constructed an object; no result was computed
naver-ai/usdm/src/decoder/voicebox/model/voicebox.py
code served (permissive licence) · get_code("0b4d4cc427b7879e")
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 an instance of its own class, SamePadLayer: the run constructed the object and did not call it, so no result was computed. 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

torch.manual_seed(0)

ARGS = {
    "num_conv_pos_embeddings": 8
}
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 SamePadLayer — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a SamePadLayer 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
SinusoidalPosEmb Ran
ran on a synthesized input · constructed an object; no result was computed
naver-ai/usdm/src/decoder/voicebox/model/voicebox.py
code served (permissive licence) · get_code("49d5a83de91032ba")
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 an instance of its own class, SinusoidalPosEmb: the run constructed the object and did not call it, so no result was computed. 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 math

torch.manual_seed(0)

# The SinusoidalPosEmb.__init__ requires a dim parameter
# The forward method expects x to be a tensor and accepts an optional scale parameter
# x should be a 1D tensor of position indices

# Create a small input tensor - batch of position indices
x = torch.tensor([0, 1, 2, 3], dtype=torch.float32)

# dim must be even, so use 4
dim = 4

# scale is optional with default 1000, but we can also test with custom value
scale = 1000

ARGS = {
    "dim": dim,
}
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 SinusoidalPosEmb — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a SinusoidalPosEmb 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
default_template Ran
ran; our contract draft was wrong, not the code
naver-ai/usdm/src/inference.py
code served (permissive licence) · get_code("1e11052145a9ed66")
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 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

torch.manual_seed(0)

# Analyze the function:
# - user_unit: used in f-string, should be a string
# - user_text: optional, used in f-string if truthy, should be a string
# - agent_text: optional, used in f-string if truthy, should be a string
# All parameters are strings used for template construction

ARGS = {
    "user_unit": "User speech and transcript",
    "user_text": "This is the user's text input",
    "agent_text": "This is the agent's response"
}
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 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
generate_bad_words_ids Ran
ran; our contract draft was wrong, not the code
naver-ai/usdm/src/inference.py
code served (permissive licence) · get_code("10df2826f0da0e19")
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 list, ndim 2, 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.

# Setup code for generate_bad_words_ids function

ARGS = {
    "start": 0,
    "end": 10,
    "exclude": [2, 5, 7]
}
Setup block drafted by
a language model, Claude Haiku 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 list, ndim 2, 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
get_slopes Ran
ran, honoured the contract we drafted
naver-ai/usdm/src/decoder/voicebox/model/voicebox.py
code served (permissive licence) · get_code("bb307ac322982165")
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 math

def get_slopes(n: int):
    def get_slopes_power_of_2(n: int):
        start = 2 ** (-2 ** (-(math.log2(n) - 3)))
        ratio = start
        return [start * ratio ** i for i in range(n)]
    if math.log2(n).is_integer():
        return get_slopes_power_of_2(n)
    else:
        closest_power_of_2 = 2 ** math.floor(math.log2(n))
        return get_slopes_power_of_2(closest_power_of_2) + get_slopes(2 * closest_power_of_2)[0::2][:n - closest_power_of_2]

ARGS = {
    "n": 8
}
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
strip_exact_multiple Ran
ran; our contract draft was wrong, not the code
naver-ai/usdm/src/inference.py
code served (permissive licence) · get_code("f18f010e48c28da6")
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 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.

# Setup code for strip_exact_multiple function

ARGS = {
    "text": "prefixHello Worldsuffix",
    "patterns": ["prefix", "suffix"]
}
Setup block drafted by
a language model, Claude Haiku 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 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
BaseModule Not yet run
unverified (harvested, no recorded run)
naver-ai/usdm/src/decoder/voicebox/model/voicebox.py
code served (permissive licence) · get_code("76cadc87e9fc55ed")
How it ran

no invocation recorded.

CFM Not yet run
unverified (harvested, no recorded run)
naver-ai/usdm/src/decoder/voicebox/model/voicebox.py
code served (permissive licence) · get_code("746887986ebb3d15")
How it ran

no invocation recorded.

EncoderLayer Not yet run
unverified (harvested, no recorded run)
naver-ai/usdm/src/decoder/voicebox/model/voicebox.py
code served (permissive licence) · get_code("0fab6f9a64d4c3c8")
How it ran

no invocation recorded.

Transformer Not yet run
unverified (harvested, no recorded run)
naver-ai/usdm/src/decoder/voicebox/model/voicebox.py
code served (permissive licence) · get_code("b7eb5907afb39920")
How it ran

no invocation recorded.

Voicebox Not yet run
unverified (harvested, no recorded run)
naver-ai/usdm/src/decoder/voicebox/model/voicebox.py
code served (permissive licence) · get_code("4dbcdd53c470b5de")
How it ran

no invocation recorded.

Also present in other papers' code

Of 15 samples lifted from this paper's linked repositories, 15 carry a function identity and 0 do not. Of the 15 distinct functions with an identity, 1 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 1 later) and 14 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
GELUActivation
brick of this paper's assembly: helper
1 other paper
1 later
2512.14391 (2025-12)
REPO: Language Models with Context Re-Positioning

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

Recent work shows promising results in expanding the capabilities of large language models (LLM) to directly understand and synthesize speech. However, an LLM-based strategy for modeling spoken dialogs remains elusive, calling for further investigation. This paper introduces an extensive speech-text LLM framework, the Unified Spoken Dialog Model (USDM), designed to generate coherent spoken responses with naturally occurring prosodic features relevant to the given input speech without relying on explicit automatic speech recognition (ASR) or text-tospeech (TTS) systems. We have verified the inclusion of prosody in speech tokens that predominantly contain semantic information and have used this foundation to construct a prosody-infused speech-text model. Additionally, we propose a generalized speech-text pretraining scheme that enhances the capture of cross-modal semantics. To construct USDM, we fine-tune our speech-text model on spoken dialog data using a multi-step spoken dialog template that stimulates the chain-ofreasoning capabilities exhibited by the underlying LLM. Automatic and human evaluations on the DailyTalk dataset demonstrate that our approach effectively generates natural-sounding spoken responses, surpassing previous and cascaded baselines. Our code and checkpoints are available at https://github.com/naverai/usdm.

For agents

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

get_harvested_code_for_paper("2402.05706")
get_code_for_paper("2402.05706")
have("2402.05706")

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

curl https://syntology.ai/api/ran/2402.05706.json

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

[![Syntology run record](https://syntology.ai/api/ran/2402.05706.svg)](https://syntology.ai/paper/2402.05706)

Connect an agent — have() is free.