Winnex AI · The manifold projector · v1.0.2

winnex-xfactor — The Engine of Manifold Projection

Built by the Winnex technical team, winnex-xfactor is the operator that projects text — coming from the winnex-nano SpectralTokenizer — onto the real manifold of a pretrained model's embeddings. Computed structurally from embed_tokens, training-free, deterministic, never by running the model.

The system

winnex-xfactor (v1.0.2) is a piece of the Winnex stack — same company, same BSL 1.1 license, same deterministic-no-black-box design principle. It is the operator that projects spectral text onto the principal subspace of a pretrained model's token embeddings.

What the algorithm does

The mathematics is classical and correct:

  1. Centers the embedding matrix E (a sample of embed_tokens).
  2. Computes the covariance C = ĒᵀĒ / V.
  3. Obtains the spectral decomposition (via power iteration + deflation, not full eig).
  4. Selects the effective rank r that captures a fraction τ of the variance.
  5. Forms the basis X = U[:, :r] and the orthogonal projector P = XXᵀ.

Mathematical properties the notebook verifies:

P² = P (idempotence) — maximum difference 0.000000
P(Pv) = Pv (projection stability)
• Implicit symmetry

This is exactly the orthogonal projection onto the principal subspace of the model's embeddings. The contribution is the fast, deterministic implementation and its integration with the rest of the stack (spectral → expand → project).

The verified benchmark

The public notebook winnex-xfactor-benchmark-manifold uses real Qwen embeddings — a serious, honest evaluation.

It loads real embeddings via winnex-nano.Safetensors + sample_embeddings (without materializing the full tensor — avoids OOM), samples 2000 rows of transformer.wte.weight (D=2048), and subsamples columns to D_sub = 32/64/128.

Main results

D_subRank (τ=0.95)Variance capturedTime
32310.97360.001 s
64600.95350.006 s
1281180.95250.042 s

Variance vs τ (D=64)

τRankVariance
0.90560.907
0.95600.954
0.99650.986

Full expand_spectral + projection pipeline works — output norm ~0.981.

The takeaway: the C++ implementation with power iteration + deflation delivers what it promises in speed for moderate ranks. D=128 → 0.042 s, D=64 → 0.006 s — fast enough for practical dimensions on a common CPU.

Assessment by component

What is solid, and the precise scope of what the operator does and does not do.

Real strengths

  • Clean, verifiable mathematics — idempotence, stable projection.
  • Training-free and structural — only looks at the embedding matrix.
  • Fast enough for D ~128–256 on a common CPU.
  • Natural integration with winnex-nano (spectral → model manifold).
  • Notebook uses real Qwen embeddings, not synthetic data.

The precise scope of the projection

What the operator does is project a vector (expanded from the spectral representation) onto the principal subspace of the token embeddings. This is not the internal residual stream, not the intermediate layer activations, and not the semantic geometry the model actually uses to generate text. It is a linear approximation of the input-embedding manifold.

Known technique, honest complexity

Power iteration + deflation is a classical numerical-linear-algebra technique (Hotelling deflation, etc.). The O(D²r) vs O(D³) gain is real when r ≪ D. At D=2048 and r ~100–200 the "~2000×" figure depends strongly on implementation and iteration count; at D=128 the notebook shows 0.042 s — perfectly acceptable, not miraculous.

Sample and downstream evidence

A 2000-embedding sample is a reasonable approximation of a ~150k-token vocabulary, not an exact decomposition of the full manifold. The reported rank (nearly D_sub) indicates the column-subsampled matrix is nearly full-rank — expected under random dimension sampling. The notebook proves the operator works (mathematics + speed + projection); downstream quality (better generation, less hallucination, better retrieval) remains a stack hypothesis to be demonstrated with standard metrics.

Role in the Winnex stack

The X-Factor closes the loop.

text → SpectralTokenizer (nano) → expand_spectral → X-Factor projector
      → vector on the model's manifold → (Madhava / StreamEngine)

It is coherent with the project philosophy — deterministic operators, training-free, mathematical proof, no black boxes. Alongside Madhava (search with bounds) and Nano (spectral tokenizer), it forms a native, auditable inference pipeline.

Verdict: winnex-xfactor is a small, well-implemented, mathematically correct library that computes the orthogonal projector onto the principal subspace of a model's token embeddings. The notebook confirms it works on real Qwen embeddings, that the projection is stable, and that power iteration + deflation is fast enough for practical dimensions.

Not a revolution in LLM representation. It is applied linear geometry, consistent with the rest of the Winnex stack. Its real utility depends on the rest of the pipeline (spectral + manifold + Madhava) delivering quality generation or retrieval — not yet demonstrated publicly with standard metrics (perplexity, MMLU, human eval).

As a piece of deterministic, training-free engineering it is solid. As "the operator that puts text into the model's latent space," the claim is read precisely: it projects onto the principal span of the input embeddings, not the deep internal geometry of the transformer.

Verifiable references

◉ PyPI — winnex-xfactor ▦ Kaggle — Manifold benchmark ← Back to the main site