Claim Routing

Claim routing decides which axis should own a new paper note or synthesis post. A paper can touch many areas, but the durable note should follow the strongest claim.

$$\text{route}(p)=\arg \underset{a}{\max }\text{claim strength}(p,a)$$

where $p$ is a paper or post candidate and $a$ is an axis such as architecture, learning method, generative modeling, computational biology, math, evaluation, systems, or agents.

First Question

Question	If Yes	Start
Is the main contribution a model structure?	architecture paper or concept update	Architecture papers
Is the main contribution an AI method claim whose type is unclear?	AI claim pattern catalog	AI paper claim patterns
Is the main contribution a supervision or training signal?	learning-method paper or concept update	Learning method papers
Is the main contribution sampling or density modeling?	generative-model paper or concept update	Generative model papers
Is the main contribution about molecules, proteins, docking, conformers, complexes, or genome-level sequence objects?	computational biology paper intake	Computational Biology paper intake
Does the paper match a repeated computational biology paper type?	claim pattern catalog	Computational Biology paper claim patterns
Does the claim require AI method, computational biology object, and Math evidence together?	cross-axis contract before writing	AI Computational Biology Math contract
Is the main contribution an equation, estimator, metric, or derivation?	formula intake or math concept update	Formula intake
Is the main contribution a benchmark or evaluation protocol?	benchmark card or evaluation concept update	Benchmark intake
Is the main contribution training, inference, reproducibility, or tooling?	systems or infra note	Systems
Is the main contribution tool-using LLM workflow behavior?	agent note	Agents

Tie Breakers

Tie	Prefer	Reason
Architecture vs learning method	learning method	if the same architecture could use the signal, the reusable idea is the objective
Architecture vs systems	systems	if speed or memory is the claim, the system boundary owns the evidence
Generative model vs molecular modeling	molecular modeling	if validity, docking, pose, affinity, or protein-ligand evaluation is central
Math vs AI method	AI method	if the formula is only the implementation of a model family
Math vs evaluation	evaluation	if the formula defines the reported metric or statistical comparison
Paper note vs concept update	concept update	if no paper-specific claim needs durable tracking
Inbox vs paper note	inbox	if metadata, public source, or relevance is still to verify

Claim Strength

Score each axis with a short note before creating a new paper page.

Axis	Evidence That It Owns The Paper
Object	new entity, modality, representation contract, or preprocessing rule
Architecture	new block, connectivity pattern, inductive bias, scaling rule, or complexity change
Learning method	new target, masking rule, contrast set, pretraining signal, preference signal, or adaptation protocol
Generative model	new likelihood, score, velocity, latent path, sampler, or validity filter
Computational biology	molecule/protein/complex object boundary, docking workflow, conformer policy, label semantics, split, or leakage issue
Math	new estimator, derivation, objective decomposition, metric definition, or uncertainty calculation
Evaluation	new benchmark, split, metric, baseline, ablation, failure taxonomy, or reproducibility evidence
Systems	compute, memory, throughput, serving, reproducibility, artifact, scaling evidence, or implementation boundary
Agents	tool contract, planning loop, memory boundary, verifier, handoff, or orchestration protocol

Routing Output

For every promoted paper or synthesis post, record:

primary axis: to verify
secondary axes: to verify
paper bucket: to verify
concept updates: to verify
paper_to_wiki_extraction: to verify
concept update contract: to verify
benchmark/evaluation update: to verify
formula update: to verify
public artifact status: to verify

Minimum Bundle By Axis

Primary Axis	Minimum Companion Notes
Architecture	architecture concept, complexity or scaling note, ablation note, evaluation risk
Learning method	objective formula, data/signal description, transfer or evaluation protocol
Generative model	probability path or sampling formula, validity metrics, failure examples
Computational biology	object contract, preprocessing/split rule, leakage check, task metric
Math	symbol definitions, sampled distribution, relation to model or metric
Evaluation	benchmark contract, metric definition, selection rule, uncertainty
Systems	artifact availability, reproducibility checklist, compute or runtime boundary
Agents	tool contract, acceptance criteria, evidence ledger, verification loop

Related

• Paper triage
• AI Computational Biology Math paper template
• Paper to wiki extraction
• Concept update contract
• Coverage matrix
• AI Computational Biology Math contract
• Scaling claim contract
• AI paper intake
• AI paper claim patterns
• Computational Biology paper intake
• Computational Biology paper claim patterns
• Formula intake
• Claim extraction
• Evidence table