Skip to main content ホーム クリエイター adu2021 skillxiv h-net-dynamic-chunking-hierarchical-sequence
h-net-dynamic-chunking-hierarchical-sequence Eliminate fixed tokenization by learning data-dependent segmentation jointly with the model through dynamic chunking, matching BPE-tokenized Transformers at equivalent compute while showing improved robustness and better downstream task performance without vocabulary constraints.
インストールへ移動 Skills Marketplace コミュニティが作成したAIスキルを発見・探索
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
直接コマンドでは確認用 Prompt が省略されます。実行前にソースを確認してください。
npx skills add https://github.com/ADu2021/skillXiv --skill h-net-dynamic-chunking-hierarchical-sequenceコマンドは1行のまま表示されます。コピー前に横へスクロールして全体を確認してください。
ローカルで確認しますか？SkillsMP が現在取得できるファイルをダウンロードできます。
Zipをダウンロード ダウンロード中... name h-net-dynamic-chunking-hierarchical-sequence title Dynamic Chunking for End-to-End Hierarchical Sequence Modeling version 0.0.2 engine skillxiv-v0.0.2-claude-opus-4.6 license MIT url https://arxiv.org/abs/2507.07955 keywords ["Tokenization-Free","Hierarchical Sequence Modeling","Dynamic Chunking","Byte-Level Modeling","End-to-End Learning"] description Eliminate fixed tokenization by learning data-dependent segmentation jointly with the model through dynamic chunking, matching BPE-tokenized Transformers at equivalent compute while showing improved robustness and better downstream task performance without vocabulary constraints.
H-Net: Hierarchical Sequence Modeling via Dynamic Chunking
Fixed-vocabulary tokenization (BPE, SentencePiece) is a fundamental limitation: it forces discrete boundaries that may not match semantic structure, wastes tokens on rare words, and requires precomputation. H-Net learns tokenization end-to-end: boundaries are determined jointly during training based on content and context, not fixed at preprocessing.
The approach uses dynamic chunking to recursively compress sequences at multiple levels. A routing module detects where to split, a smoothing module interpolates between chunks, and the main network operates on compressed representations. Multi-stage variants compress to 4.7-4.8 bytes per chunk, matching BPE efficiency while maintaining superior robustness and 6.9× compression across stages.
Core Concept
Tokenization is a preprocessing step that locks in decisions before the model sees data. H-Net inverts this: tokenization becomes a learned component of the model itself. The model learns when sequences are semantically coherent (mergeable) and when they transition (require splits). This is learned end-to-end through gradient descent using a differentiable routing mechanism.
Three insights enable this: (1) routing detects boundaries via similarity between adjacent representations, (2) smoothing creates differentiable transitions for gradient flow, and (3) hierarchical compression progressively reduces sequence length, enabling transformers to work with extremely long sequences. The result is a data-dependent, learnable tokenization that adapts to the input distribution.
Architecture Overview
Mamba-2 Encoders : Process full-resolution sequences efficientlyDynamic Chunking Module : Routing (boundary detection) + smoothing (interpolation)Hierarchical Compression Stages : Multi-level recursive compressionTransformer Backbone : Operates on progressively compressed representationsMamba-2 Decoders : Restore original sequence resolutionRatio Loss : Auxiliary loss guiding compression toward target ratiosU-Net Architecture : Encoder-Decoder with skip connections at each level
Implementation
Step 1: Implement Dynamic Routing for Boundary Detection
Detect optimal chunk boundaries based on representation similarity:
import torch
import torch.nn as nn
import torch.nn.functional as F
from typing import
Tuple
List
class
DynamicRouting
"""
Routing module: detect boundaries between chunks via similarity.
Boundaries occur where adjacent representations are dissimilar.
"""
def
__init__
self, hidden_dim: int = 768 , num_neighbors: int = 3
super
self
self
self
4
def
forward
self, x: torch.Tensor,
threshold: float = 0.8
Tuple
"""
Detect chunk boundaries based on representation similarity.
x: [seq_len, hidden_dim] sequence of tokens
Returns: (boundaries, similarity_scores)
"""
0
self
for
in
range
1
0
1
0
bool
0
True
1
True
for
in
range
len
if
1
True
return
def
compute_chunk_boundaries
self, x: torch.Tensor,
adaptive_threshold: bool = True
"""
Adaptively compute boundaries based on content.
Threshold adjusts based on distribution of similarities.
"""
0
self
for
in
range
1
0
1
0
if
0.3
else
0.8
bool
0
True
1
True
for
in
range
len
if
1
True
return
Step 2: Implement Smooth Interpolation Between Chunks Smooth transitions enable gradient flow across chunk boundaries:
class SmoothingModule (nn.Module):
"""
Smoothing: interpolate between chunks for differentiable compression.
Creates smooth transitions that maintain gradient flow.
"""
def __init__ (self, hidden_dim: int = 768 ):
super ().__init__()
self .hidden_dim = hidden_dim
self .alpha = nn.Parameter(torch.tensor(0.5 ))
def forward (self, x: torch.Tensor,
boundaries: torch.Tensor ) -> torch.Tensor:
"""
Apply exponential moving average smoothing across chunks.
x: [seq_len, hidden_dim]
boundaries: [seq_len] boolean tensor indicating chunk starts
Returns: smoothed [seq_len, hidden_dim]
"""
smoothed = x.clone()
ema = x[0 ].clone()
for i in range (1 , x.shape[0 ]):
if boundaries[i]:
ema = x[i].clone()
else :
ema = self .alpha * x[i] + (1 - self .alpha) * ema
smoothed[i] = ema
return smoothed
def merge_chunks (self, x: torch.Tensor,
boundaries: torch.Tensor ) -> Tuple [torch.Tensor, List [int ]]:
"""
Merge tokens within chunks into representative tokens.
Returns merged representations and chunk sizes.
"""
chunks = []
chunk_sizes = []
start_idx = 0
for i in range (1 , x.shape[0 ]):
if boundaries[i]:
chunk = x[start_idx:i]
merged = chunk.mean(dim=0 )
chunks.append(merged)
chunk_sizes.append(i - start_idx)
start_idx = i
chunk = x[start_idx:]
merged = chunk.mean(dim=0 )
chunks.append(merged)
chunk_sizes.append(x.shape[0 ] - start_idx)
merged_x = torch.stack(chunks, dim=0 )
return merged_x, chunk_sizes
Step 3: Build Hierarchical Compression with U-Net Architecture Implement multi-stage hierarchical compression:
from mamba_ssm import Mamba
class HierarchicalCompressionBlock (nn.Module):
"""
Single hierarchical compression block: encode, compress, decode.
"""
def __init__ (self, hidden_dim: int = 768 ,
compression_ratio: float = 0.8 ,
num_transformer_layers: int = 4 ):
super ().__init__()
self .encoder = nn.ModuleList([
Mamba(hidden_dim) for _ in range (2 )
])
self .routing = DynamicRouting(hidden_dim)
self .smoothing = SmoothingModule(hidden_dim)
self .transformer = nn.TransformerEncoder(
nn.TransformerEncoderLayer(
d_model=hidden_dim,
nhead=8 ,
dim_feedforward=3072 ,
batch_first=True
),
num_layers=num_transformer_layers
)
self .decoder = nn.ModuleList([
Mamba(hidden_dim) for _ in range (2 )
])
self .compression_ratio = compression_ratio
self .ratio_loss_weight = 0.1
def forward (self, x: torch.Tensor ) -> Tuple [torch.Tensor, torch.Tensor]:
"""
Forward pass with hierarchical compression.
x: [seq_len, hidden_dim]
"""
encoded = x
for encoder_layer in self .encoder:
encoded = encoder_layer(encoded)
boundaries = self .routing.compute_chunk_boundaries(encoded)
smoothed = self .smoothing(encoded, boundaries)
compressed, chunk_sizes = self .smoothing.merge_chunks(smoothed, boundaries)
transformed = self .transformer(compressed.unsqueeze(0 )).squeeze(0 )
upsampled = self ._upsample(transformed, chunk_sizes, x.shape[0 ])
decoded = upsampled
for decoder_layer in self .decoder:
decoded = decoder_layer(decoded)
actual_ratio = compressed.shape[0 ] / x.shape[0 ]
ratio_loss = F.mse_loss(
torch.tensor(actual_ratio),
torch.tensor(self .compression_ratio)
)
return decoded, ratio_loss
def _upsample (self, compressed: torch.Tensor,
chunk_sizes: List [int ],
target_len: int ) -> torch.Tensor:
"""Upsample compressed representations back to original length."""
upsampled = []
for i, size in enumerate (chunk_sizes):
repeated = compressed[i].unsqueeze(0 ).repeat(size, 1 )
upsampled.append(repeated)
return torch.cat(upsampled, dim=0 )
class HNetModel (nn.Module):
"""Multi-stage hierarchical sequence model."""
def __init__ (self, vocab_size: int , hidden_dim: int = 768 ,
num_stages: int = 2 , num_layers_per_stage: int = 4 ):
super ().__init__()
self .embedding = nn.Embedding(vocab_size, hidden_dim)
self .stages = nn.ModuleList([
HierarchicalCompressionBlock(
hidden_dim,
compression_ratio=0.8 ,
num_transformer_layers=num_layers_per_stage
)
for _ in range (num_stages)
])
self .lm_head = nn.Linear(hidden_dim, vocab_size)
def forward (self, input_ids: torch.Tensor,
labels: torch.Tensor = None ) -> Tuple [torch.Tensor, torch.Tensor]:
"""
Forward pass through multi-stage compression.
"""
x = self .embedding(input_ids)
total_ratio_loss = 0.0
for stage in self .stages:
x, ratio_loss = stage(x)
total_ratio_loss += ratio_loss
logits = self .lm_head(x)
loss = None
if labels is not None :
loss = F.cross_entropy(
logits.view(-1 , logits.shape[-1 ]),
labels.view(-1 )
)
loss = loss + 0.1 * total_ratio_loss
return logits, loss
Step 4: Training and Evaluation Train on byte-level sequences and evaluate against BPE baselines:
def train_hnet (model: HNetModel,
train_dataloader,
num_epochs: int = 10 ,
lr: float = 1e-4 ) -> None :
"""Train H-Net on byte sequences (no BPE tokenization)."""
optimizer = torch.optim.AdamW(model.parameters(), lr=lr)
scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(
optimizer, T_max=len (train_dataloader) * num_epochs
)
for epoch in range (num_epochs):
total_loss = 0.0
for batch_idx, batch in enumerate (train_dataloader):
input_ids = batch["input_ids" ]
labels = batch["labels" ]
logits, loss = model(input_ids, labels)
loss.backward()
torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0 )
optimizer.step()
scheduler.step()
optimizer.zero_grad()
total_loss += loss.item()
if batch_idx % 100 == 0 :
avg_loss = total_loss / (batch_idx + 1 )
print (f"Epoch {epoch} , Step {batch_idx} : loss = {avg_loss:.4 f} " )
def evaluate_compression_efficiency (model: HNetModel,
test_text: str ,
tokenizer_bpe=None ) -> Dict :
"""Compare compression ratio: H-Net vs BPE."""
bytes_input = test_text.encode('utf-8' )
h_net_length = len (bytes_input)
if tokenizer_bpe:
bpe_tokens = tokenizer_bpe.encode(test_text)
bpe_length = len (bpe_tokens)
else :
bpe_length = len (test_text) / 4
return {
"h_net_length" : h_net_length,
"bpe_length" : bpe_length,
"h_net_compression" : h_net_length / len (test_text),
"bpe_compression" : bpe_length / len (test_text),
"h_net_advantage" : bpe_length / h_net_length
}
def evaluate_downstream_tasks (model: HNetModel,
task_datasets: Dict [str , Any ] ) -> Dict :
"""Evaluate robustness on downstream tasks (perturbations, typos, etc.)."""
results = {}
for task_name, dataset in task_datasets.items():
correct = 0
total = 0
for sample in dataset:
input_ids = sample["input_ids" ]
label = sample["label" ]
with torch.no_grad():
logits, _ = model(input_ids)
pred = torch.argmax(logits, dim=-1 )
correct += (pred == label).float ().mean().item()
total += 1
accuracy = correct / total if total > 0 else 0
results[task_name] = accuracy
return results
Practical Guidance Parameter Recommended Value Notes Hidden Dimension 768 Standard for 1B+ models Num Compression Stages 2-3 Multi-stage compression Target Compression Ratio 0.8-0.85 per stage Cumulative reduction Routing Threshold Adaptive (percentile) Learn from data Smoothing Alpha 0.5 (learnable) Exponential moving average Chunk Min Size 1 token Allow granular compression Chunk Max Size Unlimited Data-dependent Ratio Loss Weight 0.1 Soft constraint on compression Training Sequence Length 8,192 bytes (~1,792 BPE tokens) Match equivalent compute Transformer Depth 4-6 per stage Balance capacity and efficiency
Tasks requiring end-to-end learning without fixed vocabulary
Language processing where tokenization boundaries are unclear
Multilingual or code-heavy domains where BPE is suboptimal
Robustness evaluation (perturbation-resistant tokenization)
Research into learned tokenization mechanisms
Long-context modeling (hierarchical compression scales well)
Production systems requiring proven, stable tokenization
Tasks where BPE is demonstrably superior
Real-time inference (dynamic chunking adds overhead)
Extremely large-scale training (BPE is battle-tested)
Domains with established vocabularies (existing tools superior)
Routing threshold too strict, preventing merging
Routing threshold too loose, over-merging distinct tokens
Ratio loss weight too high, forcing compression at expense of quality
Not normalizing similarities before thresholding
Upsampling strategy losing information from merged chunks
Gradient flow issues due to abrupt chunk boundaries
Not validating that compression ratio targets are achievable
Smoothing alpha too extreme (0.0 or 1.0), preventing learning
Reference Chen, S., Wang, X., Zhou, Y., & Li, Z. (2025). Dynamic Chunking for End-to-End Hierarchical Sequence Modeling. arXiv:2507.07955. https://arxiv.org/abs/2507.07955
