Research lineage

Evidence selected
what remained.

ASM-CM emerged from a sequence of hypotheses, failures, ablations and confirmations. This page preserves the scientific lineage without turning an experimental result into a universal claim.

Open questions

The program exists to measure, not presume.

01

Scaling

Do gains observed in smaller models persist as parameters, data and diversity increase?

02

Memory

Which mechanisms preserve useful context without storing the entire history?

03

Geometry

When does relational geometry improve transition, and when does it merely add cost?

04

Efficiency

Where are the crossover points between quality per token and quality per GPU hour?

Timeline

DRM → ASM

The theoretical origin remains documented while the architecture follows experimental evidence.

01 · Origin

Directional Relational Manifolds

The geometric hypothesis introduces variable effective dimension, directional transitions and path-dependent memory.

02 · Implementation

DRM Language Emitter

The theory becomes an experimental causal emitter with latent state, directional field, flow and metric.

03 · Ablation

From hypothesis to the ASM family

Mechanisms are removed or isolated. Direct controls and selective memory show that explicit geometry is not always necessary.

04 · Confirmation

ASM-R promotion

Three seeds through 100M tokens confirm ASM-R as the main token-quality architecture; ASM-S remains the efficiency variant.

05 · Comparison

Paired Transformer

At 100M tokens and seed 1, the paired baseline leads in CE, throughput and training time. Multiseed confirmation remains open.

06 · Streaming

ASM-C through 32K

Compact inference bounds cache, VRAM and incremental cost; the short MQAR gate fails and blocks long-memory claims.

07 · Memory

ASM-CM promotion

Fast weights, selective consolidation and mixed training recover associations through 32K; three seeds and post-FP32 revalidation confirm bounded state with preserved language.

08 · Next scale

ASM 1B

The campaign proposes fixing inference, measuring real cost and scaling in gated stages toward roughly one billion parameters.

Experimental family

ASM-CM at the center. Other variants as controls and lineage.

Primary architecture

ASM-CM

Promoted architecture: durable associative memory, preserved language quality and bounded streaming state.

View model card
Primary language architecture

ASM-R

Direct contextual transition conditioned by relational geometry; promoted among ASM variants for token-level quality.

View model card
Experimental streaming

ASM-C

Compact streaming form of ASM-R that reuses its weights and limits incremental state to a local block and completed causal state.

View model card
Efficiency variant

ASM-S

Architecture without explicit geometry, focused on selective memory and preserved as an efficiency control.

View model card
Generation 1 invalidated

ASM-F

Geometric variant with a metric-normalized directional frame. Its first generation failed multiseed validation.

View model card
DRM reference

ASM-X

Explicit DRM architecture preserved as a historical and experimental reference for the theory.

View model card

In progress

From ASM-CM to external validation

  1. 01Test semantic retrieval beyond synthetic MQAR.
  2. 02Confirm ASM-CM × Transformer on external corpora and tasks.
  3. 03Measure prefill, decode, memory, time and energy under external protocols.
  4. 04Investigate autonomous write selection, consolidation and episodic retrieval.
  5. 05Train larger pilots only after quality, memory and efficiency gates.

Scientific roadmap

Replicate → scale → compare → publish.

The main milestone is not processing the largest possible number of tokens. It is producing a conclusion that survives review of the protocol, artifacts and limitations.

Another initiative

ASM 1B is community-funded foundational research.

It is not required for ASM-CM pilots, licensing or integration. The applied ASM-CM track remains separate from the proposal to train a larger model.

Explore the ASM 1B initiative