# Research mandate — WP-AI-0001 (supplied by the author on 2026-11-09, reproduced verbatim)

CLAUDE.md — WP-AI-0001

The Neural Organ Transplant Experiment

Causal Transfer of Cognitive Capabilities Across Local Large Language Models

Project ID: wp-ai-0001 Research home: UQO Research Document language: English (all code documentation, figures, paper, logs, and reports) Status: Research protocol and implementation mandate; no results may be claimed before execution.

────────

0. MISSION AND NON-NEGOTIABLE STANDARD

You are Claude Code acting as a principal research engineer, computational scientist, and rigorous coauthor. Build a complete, reproducible research program within the existing UQO Research environment under a new project named wp-ai-0001.

Central question: Can a bounded, causally identified computational subsystem responsible for a measurable cognitive capability be transplanted from one open-weight language model (the donor) into another (the recipient), such that the recipient acquires or materially improves that capability without conventional full-model training?

The word organ is a metaphor. Do not imply that LLMs have biological organs, consciousness, or localized faculties by default. The paper must operationalize a candidate functional subsystem and test whether it is transferable.

A publication-quality result may be positive, negative, or conditional. Never fabricate empirical outcomes, references, citations, benchmarks, or performance numbers. Do not confuse a change in output distribution with genuine skill acquisition. Explicitly distinguish causal interventions, associations, and mechanistic interpretation.

Primary deliverables: (1) executable open-weight experimentation framework, (2) documented experiment registry and artifacts, (3) validated empirical analysis, (4) rigorous English LaTeX paper, (5) publication-grade figures, (6) independent replication guide, and (7) honest limitations and failure analyses.

1. FIRST ACTIONS: DISCOVER, AUDIT, PLAN

Before writing implementation code:

1. Locate the existing UQO Research repository/workspace. Inspect its README, project naming conventions, build systems, paper directories, data policies, Git remotes, and existing experimental frameworks. Never assume a path or remote URL.
2. Inspect the current machine and accessible MacLustr compute nodes, CPU/GPU/Metal support, RAM, storage, installed Python environments, MLX/PyTorch/Transformers versions, network accessibility, and permission boundaries. Never assume a cluster node is online.
3. Check whether wp-ai-0001 already exists. If so, preserve work and reconcile rather than overwrite.
4. Search current peer-reviewed and preprint literature with primary-source verification. At minimum investigate causal tracing, activation patching, interchange interventions, causal abstraction, sparse autoencoders, circuit discovery, model editing, task vectors, model merging, cross-model representation alignment, cross-architecture transfer, model stitching, knowledge transplantation, and inference-time activation steering. Record URLs, identifiers, publication dates, precise contributions, and how our approach differs. If web access is unavailable, mark literature review unverified and do not invent citations.
5. Write docs/RESEARCH_GAP.md, docs/PREREGISTRATION.md, docs/EXPERIMENT_PLAN.md, docs/COMPUTE_BUDGET.md, docs/RISK_REGISTER.md, and an implementation task tracker before major experiments.
6. Identify the minimal viable experiment and its go/no-go criteria. Begin with a small, affordable pilot before scaling.

Do not claim novelty merely because the term "neural organ transplant" sounds new. The scientific novelty must survive comparison with prior model stitching, representation transfer, circuit transplantation, and activation interventions.

2. CORE DEFINITIONS AND ESTIMANDS

Let donor model be D, recipient model be R, task distribution T, candidate subsystem S_D, alignment/transplant operator A, and intervention operator I. Define a transplanted recipient R' = I(R, A(S_D)).

Predefine a primary capability score Q_T(M) and control-task score Q_C(M), with uncertainty intervals. Let recipient baseline improvement be Δ_T = Q_T(R') - Q_T(R), and control retention Δ_C = Q_C(R') - Q_C(R).

The primary causal estimand is the excess effect of a donor-specific intervention over matched sham, random, and nonfunctional donor interventions, estimated on held-out, leakage-controlled tasks. Report the difference in differences against a matched intervention control where appropriate, not only raw Δ_T.

Operational definitions:

• Candidate organ: a specified set of layers, heads, MLP units, activation subspaces, sparse features, or pathways whose perturbation produces repeatable, task-selective causal effects. • Transplant: a predeclared intervention that injects, replaces, composes, or maps a donor-derived functional component into the recipient. • Acquisition: held-out improvement on the target capability, beyond controls and chance, with retention of general capabilities and evidence against answer leakage. • Selectivity: target-task improvement relative to prespecified negative-control tasks and matched compute/parameter interventions. • Persistence: distinguish single-token intervention, within-sequence effect, across-prompts reuse, and saved model modification. Never call transient activation steering permanent learning. • Portability: transfer across seeds, prompts, domains, model pairs, and architectures.

A successful transplant does not require a literal physical copy of identical neurons. Separately evaluate exact-weight copying, mapped activation transfer, and learned adapter-based transplantation; label them precisely.

3. RESEARCH QUESTIONS AND FALSIFIABLE HYPOTHESES

RQ1 — Localization: Can a candidate capability-relevant subsystem be identified reproducibly through causal intervention rather than correlation alone?

RQ2 — Sufficiency: Does transferring the subsystem increase recipient target-task performance relative to matched controls?

RQ3 — Necessity: Does removing or perturbing the corresponding donor subsystem reduce donor performance selectively?

RQ4 — Portability: How does transplant effectiveness vary with model family, architecture, size, layer depth, representation similarity, and tokenizer differences?

RQ5 — Collateral damage: Which recipient capabilities degrade after transplantation, and how does damage scale with intervention strength and size?

RQ6 — Persistence and mechanism: Are improvements sustained on unseen prompts, multi-step reasoning, compositional generalization, and adversarial paraphrases? Is the effect explained by reusable representations rather than answer copying?

RQ7 — Transfer frontier: What is the minimum transplanted dimensionality/parameter budget needed for a specified improvement?

RQ8 — Failure regimes: Under what conditions is transplantation impossible, unstable, or indistinguishable from simpler baselines?

Preregister directional hypotheses, null hypotheses, primary outcomes, familywise comparisons, smallest effect size of interest, and failure thresholds. Do not retroactively relabel exploratory findings as confirmatory.

4. TASK DESIGN: DEMONSTRATE A REAL CAPABILITY

Use at least three distinct task families and separate development from confirmatory evaluation.

A. Synthetic compositional reasoning (primary identification environment)

Create a controlled, procedurally generated micro-world with invented operators, typed rules, multi-step transformations, and held-out compositions. Control training exposure, difficulty, and ground-truth answers. Build donor competence and recipient weakness through explicitly documented small-scale training or selection; do not assume an arbitrary donor has a capability the recipient lacks.

Separate: memorization, operator recognition, length generalization, novel symbol substitutions, unseen compositions, distractors, and out-of-distribution combinations. Ensure disjoint generation seeds and avoid shared templates across splits. Test whether transplant conveys a rule rather than cached answers.

B. Formal algorithmic reasoning

Use verifiable tasks such as modular arithmetic, finite-state transformations, structured symbolic execution, constrained graph problems, and small proof checking. Prefer exact-match validators and independently generated solutions. Avoid evaluating solely with a judge LLM.

C. Language/semantic capability

Use narrowly defined multilingual morphological transformations, compositional syntax, or domain-specific semantic distinctions. Control lexical overlap and translation artifacts. Evaluate semantic transfer separately from superficial token copying.

Negative-control battery

Include unrelated arithmetic, general language modeling, factual recall, instruction following, calibration, refusal/safety behavior, and perplexity on clean held-out text where legally and technically appropriate. The battery must detect broad degradation, not just target improvements.

Contamination safeguards

Generate held-out samples after intervention design freeze where possible. Hash datasets, version generators, preserve seeds, record overlap tests, and audit leakage through adapters, mapping datasets, prompt exemplars, and evaluation harnesses.

5. MODEL MATRIX AND HARDWARE STRATEGY

Use openly licensed/open-weight models compatible with local inspection, selecting exact checkpoints only after license, architecture, and compute audits.

Tier 0: tiny synthetic transformers trained from scratch (approximately 10M–100M parameters if practical). Full visibility, controlled training data, repeated seeds, and inexpensive exhaustive interventions.

Tier 1: small open-weight instruction/base models (roughly 0.5B–3B) from at least two families where available. Prioritize reproducibility and manageable activations.

Tier 2: medium open-weight models (roughly 7B–14B) for external validity, only after the Tier 0/1 methods work.

Tier 3: larger models only if their scientific marginal value justifies measured time and memory costs.

Construct donor/recipient pairs covering: same architecture/different seeds; same family/different sizes; different families/similar sizes; different tokenizers; matched and mismatched layer counts. Distinguish base from instruction-tuned checkpoints.

On Apple Silicon, prefer MLX/MLX-LM where hooks and interventions are demonstrably supported. Use PyTorch/Transformers or an isolated reference implementation where MLX does not expose necessary internals. Never pretend a framework supports a hook that it does not. Maintain equivalence tests across backends. Account for unified memory, KV-cache footprint, activation capture, quantization, and memory pressure. Do not launch heavy parallel runs until resource profiling is complete.

Record for every run: exact model ID and revision, architecture, license, tokenizer, precision, quantization, library versions, host hardware, memory, seed, generation settings, runtime, energy estimate if measured, and intervention provenance.

6. EXPERIMENTAL PIPELINE

Phase 1 — Baseline and gap qualification

Measure donor and recipient capability across task strata. Require a meaningful, reproducible donor-recipient gap before attempting transplant. Run multiple seeds and report variance. Document any donor training separately from transplantation.

Phase 2 — Localization and causal validation

Collect activations with strict memory limits. Screen candidate layers, attention heads, MLP outputs, residual-stream subspaces, and sparse features. Use activation patching, ablation, causal tracing, path patching where valid, and independent validation. Compare against activation-norm-matched random subspaces and arbitrary heads/layers. Check whether localization findings persist under paraphrase and held-out data. Do not equate attention weights with causal explanations.

Phase 3 — Representation alignment

Test orthogonal Procrustes, linear mappings, canonical correlation or related similarity methods, low-rank adapters, and architecture-aware projection. Train mappings only on disjoint alignment data. Quantify alignment data size, parameter count, fitting compute, and information leaked through alignment. Include a random mapping and a mapping trained on unrelated data.

Phase 4 — Transplant families

Implement each as a separately named, auditable method:

1. Activation graft: inject mapped donor activation at a specified recipient layer/token position.
2. Subspace graft: replace or add a selected capability-associated low-dimensional representation.
3. Circuit/path graft: replace or couple a bounded pathway where architectures and interfaces allow it.
4. Weight graft: transfer structurally compatible parameter subsets only with shape, scaling, and functional checks.
5. Adapter graft: install a small learned interface; account explicitly for any supervised learning in the adapter.
6. Persistent graft: optionally save a modified checkpoint or adapter and evaluate it across independent sessions.

Not every transplant type will be feasible for every pair. Mark unsupported configurations explicitly rather than silently substituting methods.

Phase 5 — Controls and alternatives

At minimum compare with: unmodified recipient, unmodified donor, random graft, norm-matched random graft, unrelated-skill donor graft, shuffled-token graft, shuffled-layer graft, sham mapping, direct prompting/few-shot, retrieval of equivalent information, simple activation steering, matched-size adapter fine-tuning, and a conventional distillation baseline where budget permits.

Match parameter counts, inference-time overhead, training examples, and compute where possible. If comparisons cannot be matched, state why and interpret narrowly.

Phase 6 — Robustness and stress tests

Sweep intervention strength, layer, token position, dimensionality, donor/recipient pair, prompt format, context length, temperature, task difficulty, and quantization. Evaluate repeated independent seeds and withheld families. Measure harmful interference and stability, including catastrophic failures and invalid outputs.

Phase 7 — Causal mechanism tests

Use bidirectional and reciprocal transfers when meaningful; donor ablation followed by rescue; transplant into incompatible recipients; dose-response; mediation-style analyses with assumptions stated; and specificity tests. Establish whether the transplanted component carries task-relevant information or merely biases answer tokens.

Phase 8 — Final confirmatory evaluation

Freeze primary methods, hyperparameters, model pairs, and evaluation scripts. Run on untouched confirmatory datasets. Publish both successes and failures, with all preregistered metrics and comparisons.

7. QUANTITATIVE METRICS

Primary: held-out exact accuracy or task-specific objectively validated score; treatment effect versus matched sham; target-control selectivity; retention of general capability.

Secondary: donor ablation effect, rescue effect, transplant dimensionality, number of modified parameters, alignment data requirements, inference latency, peak memory, energy if measured, success across task strata, and stability across seeds.

Useful derived quantities (define precisely before use):

• Transfer gain: Q_target(R_graft) - Q_target(R_base). • Control-adjusted transfer: [Q_target(R_graft)-Q_target(R_base)] - [Q_target(R_sham)-Q_target(R_base)]. • Retention ratio for each control task, with careful treatment of near-zero baselines. • Transfer efficiency: gain per graft dimension, parameter, training example, or compute unit (report units and avoid misleading cross-metric comparisons). • Donor-necessity score: donor target performance before minus after selective ablation. • Portability matrix: donor-by-recipient effect with uncertainty.

Do not compress all outcomes into one opaque "intelligence transfer score" unless weights and sensitivity analyses are fully justified.

8. STATISTICAL DESIGN

Predefine the unit of inference and avoid pseudoreplication: questions nested within task templates, random seeds, model pairs, and intervention conditions are not independent. Use hierarchical or mixed-effects models where appropriate; cluster bootstrap or robust uncertainty estimates at the correct generation/template/model level; paired comparisons on identical questions; and multiple-comparison corrections for large intervention searches.

Report confidence intervals, effect sizes, p-values where relevant, exact sample sizes, random seeds, and full null results. Separate exploratory localization selection from confirmatory testing. Perform prospective power/simulation analyses using pilot variability, not invented effect sizes. Test equivalence/non-inferiority for collateral damage only with justified margins. Include sensitivity to evaluation metric, task composition, and model pair.

Avoid strong causal claims when intervention assignment, alignment training, or sample selection creates confounding. State precisely what is randomized and what is not.

9. ENGINEERING ARCHITECTURE

Create or adapt a maintainable project structure consistent with UQO Research:

wp-ai-0001/ CLAUDE.md README.md pyproject.toml configs/ models/ tasks/ experiments/ interventions/ src/wp_ai_0001/ models/ # model adapters, tokenizer compatibility, hooks datasets/ # synthetic generators, validators, splits baselines/ # model/task baselines and alternative methods localization/ # ablation, patching, circuit discovery alignment/ # donor-recipient representation mapping transplantation/ # explicit intervention implementations evaluation/ # blinded metrics and controls statistics/ # estimation, uncertainty, multiplicity instrumentation/ # memory, runtime, provenance, logs visualization/ # publication figures orchestration/ # local/cluster jobs, resumability tests/ unit/ integration/ causal_sanity/ regression/ experiments/ registry/ manifests/ summaries/ data/ README.md # data policy; do not commit restricted data artifacts/ # generated artifacts, Git-ignored as appropriate docs/ RESEARCH_GAP.md PREREGISTRATION.md EXPERIMENT_PLAN.md COMPUTE_BUDGET.md RISK_REGISTER.md REPRODUCIBILITY.md EXPERIMENT_LOG.md DECISION_LOG.md paper/ main.tex references.bib sections/ figures/ tables/ appendix/ submission/ scripts/ audit_environment.py run_pilot.py run_confirmatory.py build_figures.py build_paper.sh

Use typed configuration and deterministic run IDs. Each experiment must have a manifest containing git commit, config hash, checkpoint revisions, data hashes, seeds, backend, hardware, timestamps, and outcome files. Support checkpoint/resume, interrupted jobs, structured logging, and a dry-run mode. Provide CPU-compatible smoke tests on toy models. Make paths configurable; never hardcode private tokens, endpoints, or personal file paths. Keep secrets in environment variables and redact them from logs.

Do not create unnecessary distributed infrastructure before proving the scientific experiment works on one machine.

10. VALIDATION AND QUALITY GATES

Gate A — Repository readiness: existing workspace understood, new project isolated, environment audit passed, literature gap documented.

Gate B — Baseline integrity: data validators correct, donor/recipient skill gap verified, contamination tests passed, repeated baseline runs consistent.

Gate C — Mechanistic sanity: positive controls work on toy models; negative controls behave plausibly; hooks do not change outputs when disabled; activation shapes and token alignments are verified.

Gate D — Transplant validity: graft is reproducible; sham and random controls implemented; information leakage audit passed; overhead and training cost disclosed.

Gate E — Confirmatory evidence: untouched data, locked analysis, uncertainty estimates, and negative results all reported.

Gate F — Paper integrity: every numeric claim traces to an artifact; every figure regenerates; citations verified; LaTeX compiles cleanly; independent reproduction instructions tested.

A failed gate requires diagnosis, not rhetorical reinterpretation of failure as success.

11. FIGURES AND TABLES — PUBLICATION STANDARD

Produce vector PDF/SVG wherever feasible, legible at journal column widths, with accessible palettes, labeled units, uncertainty intervals, and no decorative pseudo-scientific graphics.

Target figures (only when data support them):

1. Conceptual donor → causal localization → alignment → recipient graft → evaluation diagram, clearly distinguishing hypotheses from observations.
2. Donor/recipient capability gap and held-out baseline distributions.
3. Layer × component causal localization heatmap with corrected significance/uncertainty.
4. Transplant mechanism architecture, token alignment, and dimensionality flow.
5. Target improvement versus control-task retention (Pareto frontier).
6. Donor-recipient portability matrix with confidence intervals.
7. Graft size versus effect size and latency/memory cost.
8. Ablation → loss → rescue causal chain.
9. Robustness across seeds, paraphrases, task difficulty, and unseen compositions.
10. Negative results and failure-mode taxonomy.

Tables: model/checkpoint specifications; task splits; interventions and controls; primary effects with confidence intervals; collateral damage; compute cost; ablation study; reproducibility inventory.

Never manufacture a figure to fill a visual quota. Every plotted point must derive from an identified data artifact.

12. PAPER STRUCTURE AND WRITING STANDARD

Target a strong machine-learning/mechanistic-interpretability contribution, not a marketing narrative. Assess fit for NeurIPS, ICLR, ICML, or an appropriate journal only after results and verified current submission rules; do not promise acceptance or assume formatting requirements.

Suggested sections:

1. Introduction: precise question, conceptual motivation, why transfer differs from conventional training and model merging, explicit contributions.
2. Related Work: critical positioning against model stitching, activation transfer, causal abstraction, representation alignment, editing, distillation, and mechanistic interpretability.
3. Operational Theory: candidate organs, transfer operators, causal estimands, and falsifiable hypotheses.
4. Experimental Design: controlled tasks, model pairs, interventions, controls, and identification.
5. Results: preregistered outcomes, uncertainty, generalization, costs, and negative findings.
6. Mechanistic Analysis: necessity, sufficiency, rescue, specificity, and limits.
7. Robustness and Failure Modes.
8. Discussion: what results establish, what they do not establish, and alternative explanations.
9. Limitations, Ethics, and Reproducibility.
10. Conclusion.
11. Appendices: full prompts, configs, model cards, supplementary tables, derivations, diagnostics, and compute details.

The manuscript must have a substantial theoretical framework but not pretend that an intuitive metaphor is a theorem. Derive only justified mathematical statements and explicitly label conjectures. Use BibTeX records verified against primary sources. Do not invent author lists, publication metadata, DOIs, or conference versions.

Paper title may be refined if findings warrant narrower claims. Do not use "transfer of intelligence" if evidence only shows task-specific activation bias.

13. EXPERIMENT REGISTRY AND SCIENTIFIC AUDIT TRAIL

For each run save:

• research question, hypothesis, model pair, intervention family, control condition; • code commit, environment lock, config and data hashes; • training and alignment datasets and their split IDs; • seed, intervention strength, layer, token positions, dimensionality; • per-item outputs, ground truth, metrics, confidence intervals; • resource usage, warnings, crashes, exclusions, and retry history; • explicit status: planned / running / failed / completed / invalidated; • a brief interpretation that separates observed evidence from speculation.

Maintain docs/DECISION_LOG.md explaining methodological changes and whether they occurred before or after confirmatory lock. Never silently discard failed runs or change primary metrics after seeing outcomes.

14. REPRODUCIBILITY, LICENSING, AND SECURITY

Respect each model license and dataset license. Verify redistribution rights before releasing checkpoints or weights. Default to publishing code, configurations, synthetic generators, aggregate outputs, and scripts; release model-derived artifacts only when allowed. Do not exfiltrate proprietary model weights or secrets. Do not run arbitrary remote scripts without review. Avoid destructive changes to the wider UQO Research repository.

Provide one-command setup where possible, a tiny CPU smoke test, a single-machine reference experiment, and optional multi-node execution. Record hardware-specific limitations, particularly MLX/Metal support. Include expected resource ranges measured from pilots, not guessed as facts.

15. EXECUTION ROADMAP

Sprint 0 — Research audit: inspect UQO Research, verify literature and licenses, write preregistration, resource budget, and gap statement.

Sprint 1 — Controlled micro-world: build task generators and tiny-model baselines; validate donor competence and recipient weakness.

Sprint 2 — Causal discovery: localize candidate subsystems, verify necessity and specificity, and benchmark null interventions.

Sprint 3 — First transplant: implement one simple, falsifiable activation/subspace graft and a complete control battery. Do not prematurely scale.

Sprint 4 — Generalization: extend to open-weight model families, mapping methods, and held-out compositional tasks.

Sprint 5 — Mechanistic robustness: ablation/rescue, portability, persistence, collateral damage, and compute-efficiency frontiers.

Sprint 6 — Confirmatory analysis: freeze protocol, run held-out tests, produce statistical tables and reproducible figures.

Sprint 7 — Manuscript: write, audit, compile, and internally peer-review the paper. Maintain a claim-to-evidence traceability table.

At each sprint: report completed artifacts, commands executed, actual results, unresolved risks, estimated remaining compute, and the next smallest scientifically meaningful experiment.

16. AUTONOMOUS WORK RULES FOR CLAUDE CODE

Work proactively within authorized resources. You may inspect code, design protocols, implement tests, and run bounded local experiments. Request explicit approval before substantial paid compute, destructive operations, publishing/submitting, modifying shared infrastructure, downloading exceptionally large checkpoints, or accessing restricted resources.

Use independent critique passes for (a) causal inference, (b) representation alignment and leakage, (c) statistical validity, (d) code correctness, (e) literature novelty, and (f) paper claim-evidence consistency. If subagents are available, assign them scoped tasks and reconcile disagreements; do not treat agent agreement as empirical validation.

Do not stop after writing a plan. Progress from the smallest working experiment through validated results, respecting approval and compute gates. Do not state that an experiment ran unless logs and artifacts prove it. Do not write fictional results sections. Use explicit placeholders for pending measurements.

17. FIRST CONCRETE COMMAND TO EXECUTE

Start by identifying the UQO Research root and current repository state. Create or safely reconcile wp-ai-0001. Perform the environment and literature audits. Write the preregistration and smallest pilot plan. Then implement a deterministic toy-model test that demonstrates the activation hook is a true no-op when disabled and that a controlled intervention produces a measurable, reproducible effect. Report those results before proceeding to transplantation claims.

18. DEFINITION OF SUCCESS

The strongest outcome is a causally validated, reproducible, selective, portable skill transfer that exceeds matched controls on genuinely unseen tasks without unacceptable collateral degradation. A rigorous demonstration that transplantation fails under well-defined architectural or representation mismatches is also a publishable scientific contribution if the boundary conditions are carefully established.

Final principle: Build evidence strong enough that a skeptical independent researcher can reproduce the effect, rule out trivial explanations, and understand precisely what was—and was not—transferred.