emergent-semantics-model-1024-bit-335m

Emergent Semantics — Model_1024_BIT (335M)

This repository provides Model_1024_BIT (335M) — an ablation model from the paper:

📚 Paper (Emergent Semantics Beyond Token Embeddings: Transformer LMs with Frozen Visual Unicode Representations) -

📚 Paper (Growing Transformers: Modular Composition and Layer-wise Expansion on a Frozen Substrate) -

📚 Blog Article

This checkpoint is designed to test whether Transformer LMs can learn strong internal semantic structure even when the input embedding space is frozen, non-semantic, and random.

---

What this ablation is

Model_1024_BIT uses a frozen embedding table where:

• n_embed = 1024 (embedding dimensionality equals d_model)
• Each token embedding is a binary vector (0/1 components)
• Vectors are randomly generated, but we guarantee a unique ID per token (no collisions) by construction
• The embedding weights are frozen (requires_grad=False) for the entire training run

This makes the embedding layer a fixed “token identifier substrate” rather than a learned semantic space.

---

Relation to other models in the collection

• Compared to Model_UNI_GLYPH (335M):

  • Same Transformer backbone (d_model=1024, 16 layers, 32 heads, RoPE, GELU)
  • Same embedding dimensionality (n_embed=1024)
  • Difference is only the embedding initialization:
    • UNI_GLYPH: deterministic glyph bitmap → PCA → normalization
    • 1024_BIT: random binary vectors, unique per token, frozen

• Compared to Model_unfrozen (335M):

  • Same architecture
  • Here embeddings are frozen and random; in the baseline they are trainable

Because n_embed=1024, this model is in the same parameter-count class (~335M) as UNI_GLYPH and the unfrozen baseline.

---

Model summary

• Architecture: decoder-only Transformer (GPT-like)
• Hidden size (d_model): 1024
• Layers: 16
• Heads: 32
• Positional encoding: rotary embeddings
• Activation: GELU
• Vocabulary size: 65,536
• Tokenizer: Bochkov/bvv241-2-3 compatible
• Input embeddings: frozen, random binary, n_embed=1024 (no collisions)
• Output head: not tied to the input embeddings (trained separately)

---

Files in this repo (embedding reference)

This repo includes the frozen embedding content used by the model (for reproducibility/inspection).

If you also publish a plaintext dump similar to the 16-bit model (recommended), add a link here, e.g.:

• embeddings.txt (reference mapping token → 1024-bit vector):
  https://huggingface.co/Bochkov/emergent-semantics-model-1024-bit-335m/blob/main/embeddings.txt

Note: Keeping the frozen embedding tensor (and optionally a readable dump) inside the model repo makes the checkpoint self-contained and avoids ambiguity if someone loads the tokenizer from elsewhere.

---

Tokenizer

The intended tokenizer is bvv241-2-3:

• https://huggingface.co/Bochkov/bvv241-2-3

You can load the tokenizer either from this model repo (if included) or from the standalone tokenizer repo. The key requirement is exact vocab alignment.

---

How to use (Transformers)

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Bochkov/emergent-semantics-model-1024-bit-335m")
model = AutoModelForCausalLM.from_pretrained("Bochkov/emergent-semantics-model-1024-bit-335m", trust_remote_code=True).to('cuda')

inputs = torch.tensor([tokenizer.encode("Question: What is the capital of Japan?\nAnswer:")], dtype=torch.long, device='cuda')

outputs = model.generate(
    inputs, 
    max_new_tokens=10,
    do_sample=False
)
print(tokenizer.decode(outputs[0].tolist()))

#Question: What is the capital of Japan?
#Answer:Takayama

Verify the 1024-bit frozen binary embeddings (sanity check)

The model uses a frozen nn.Embedding(vocab_size=65536, n_embed=1024) whose values are strictly binary (0/1).

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

repo_id = "Bochkov/emergent-semantics-model-1024-bit-335m"

tokenizer = AutoTokenizer.from_pretrained(repo_id)
model = AutoModelForCausalLM.from_pretrained(repo_id, trust_remote_code=True)
model.eval()

print("vocab_size:", tokenizer.vocab_size)
print("config:", {k: getattr(model.config, k) for k in ["vocab_size", "n_embed", "d_model", "n_layer", "n_head", "scale"]})

# --- 1) Show embedding matrix shape (should be 65536 x 1024) ---
W = model.token_embeddings.weight.detach().cpu()
print("token_embeddings.weight shape:", tuple(W.shape))  # (65536, 1024)

# --- 2) Tokenize 'A' and show its token id  ---
text = "A"
ids = tokenizer.encode(text, add_special_tokens=False)
tokens = tokenizer.convert_ids_to_tokens(ids)

print(f"text={text!r}")
print("ids:", ids)
print("tokens:", tokens)

tid = ids[0]

# --- 3) Print the 256 dim vector and verify it is binary (0/1) ---
e1024= W[tid]  # shape: (1024)
print("1024-dim embedding for token id", tid, ":", e1024.tolist())

uniq = torch.unique(e1024)
print("unique values in e1024", uniq.tolist())

is_binary = torch.all((e1024== 0) | (e1024== 1)).item()
print("is strictly binary (0/1):", is_binary)

# --- 4) Global check: all embedding weights are exactly 0/1 ---
is_binary_global = torch.all((W == 0) | (W == 1)).item()
num_non_binary = torch.numel(W) - torch.sum((W == 0) | (W == 1)).item()
print("is binary globally (0/1):", is_binary_global)
print("non-binary entries:", int(num_non_binary))

Expected output highlights (example):

• vocab_size: 65536
• config: {'vocab_size': 65536, 'n_embed': 1024, 'd_model': 1024, 'n_layer': 16, 'n_head': 32, 'scale': 1}
• token_embeddings.weight shape: (65536, 1024)
• text='A'
• ids: [65]
• tokens: ['A']
• 1024-dim embedding for token id 65 : [1.0, 0.0, 0.0, 0.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 0.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 1.0, 0.0, 1.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 1.0, 0.0, 0.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 1.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 1.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 1.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 0.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 1.0, 0.0, 1.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 1.0, 0.0, 0.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 1.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 1.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 1.0, 1.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 1.0, 0.0, 1.0, 0.0, 0.0, 1.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 1.0, 0.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 1.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 1.0, 1.0, 0.0, 1.0, 0.0, 1.0]
• unique values in e1024 [0.0, 1.0]
• is strictly binary (0/1): True
• is binary globally (0/1): True
• non-binary entries: 0

---

Intended use

Research-only checkpoint intended for:

• Studying emergent semantics when the embedding layer is a frozen random codebook
• Comparing glyph-based vs random frozen embeddings under identical training conditions
• Controlled interpretability experiments and ablations on embedding initialization/freezing

Not intended for production deployment (no safety/instruction tuning).

---

Related links

• Model collection (paper artifacts):
  https://huggingface.co/collections/Bochkov/emergent-semantics-beyond-token-embeddings
• UNI_GLYPH model (frozen visual glyph embeddings):
  https://huggingface.co/Bochkov/emergent-semantics-model-uni-glyph-335m
• Tokenizer:
  https://huggingface.co/Bochkov/bvv241-2-3
• Code (GitHub):
  https://github.com/AVBochkov/Embeddings

---

🧑‍🔬 Citation & Concept

If you use this model or the underlying concepts in your research, please cite our work:

@article{
      bochkov2025emergent,
      title={Emergent Semantics Beyond Token Embeddings: Transformer {LM}s with Frozen Visual Unicode Representations},
      author={Andrey Bochkov},
      journal={Transactions on Machine Learning Research},
      issn={2835-8856},
      year={2025},
      url={https://openreview.net/forum?id=Odh8IynO1o},
      note={}
}
@misc{bochkov2025growingtransformersmodularcomposition,
      title={Growing Transformers: Modular Composition and Layer-wise Expansion on a Frozen Substrate}, 
      author={A. Bochkov},
      year={2025},
      eprint={2507.07129},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2507.07129}, 
}