03 · SCoTD + CoTD — Symbolic Reasoning

SCoTD + CoTD

The synthesis stage: teacher rationales are decomposed into explicit symbolic steps, and those reasoning chains — not raw answers — become the training data.

● available now — the symbolic decomposition schema: premise → rule → constraint → cross-reference → conclusion

Roadmap: AutoResearch-evolved CoT templates per failure analysis

The technical how — reasoning, decomposed

Stage 03 of the Nucleus pipeline (wired end-to-end today; production-scale teacher-backed generation is the aim). The Synthesizer takes the knowledge instances surfaced by extraction (kice-tice) and asks the teacher model not for answers but for reasoning, then decomposes each rationale into five explicit, individually traceable symbolic steps: premise identification → rule application → constraint checking → cross-reference validation → conclusion derivation. Each step is a discrete unit (a StepType in the engine), so reasoning transfers to the student structurally — as inspectable step chains — rather than as imitated token patterns. Following the SCoTD finding that single-chain distillation underperforms, multiple reasoning chains are sampled per instance (SCOTT-style self-consistency guards faithfulness downstream at certification). The division of labor: extraction finds the what, the teacher generates the how, the student internalizes the how. The decomposed chains are written atomically to the training set that sft-raft consumes; AutoResearch (autoresearch) evolves the step templates between cycles based on failure analysis from the swarm.

Five explicit steps — every rationale, every time

Premise

identification — what is actually given

Rule

application — which rule governs the case

Constraint

checking — what bounds the rule here

Cross-reference

validation — does the corpus agree

Conclusion

derivation — the answer, earned

Each step is a discrete, individually traceable unit — a StepType in the engine — which is the whole point: reasoning transfers to the student structurally, as inspectable step chains, not as imitated token patterns. An answer is only as good as the chain that derives it, and a chain can be checked link by link.

SCoTD

Symbolic Chain-of-Thought Distillation: train a small student on many sampled teacher rationales, so models far below the CoT-emergence scale still learn to reason step-by-step.

Symbolic Chain-of-Thought Distillation (Li, Hessel, Yu, Ren, Chang, Choi — ACL 2023, arXiv:2306.14050) is the technique Stage 03 is named for. Chain-of-thought gains were thought to emerge only in very large models; SCoTD shows that students orders of magnitude smaller (125M–1.3B in the paper) benefit when trained on rationalizations sampled from a much larger teacher — with sampling multiple chains per instance a load-bearing detail, since single-chain distillation underperforms. Nucleus adopts both the core move (the rationale, not the answer, is the supervision signal) and the sampling discipline (SCOTD_SAMPLES_PER_INSTANCE), and pairs it with SCOTT-style self-consistency (arXiv:2305.01879) so the distilled rationales faithfully justify the conclusions rather than merely decorating them.

In the World →

AutoResearch

The meta-agent that evolves Stage 03's reasoning-step templates — and the extraction rubrics upstream — from failure analysis, cycle over cycle.

AutoResearch is the in-house meta-agent that closes the pipeline's improvement loop. In Stage 03 it evolves the CoT step templates: when the adversarial loop (swarm) finds the student failing a class of probes, AutoResearch traces which decomposition step went weak — premises misidentified, constraints skipped, cross-references absent — and rewrites the templates the Synthesizer uses on the next cycle. The same agent re-tunes extraction upstream (kice-tice quality scores, source reweighting, keyword evolution, cross-layer promotion), so failures propagate fixes through re-extract → re-synthesize → re-train. Status seam: AutoResearch is wired into the pipeline today but falls back to an empty rubric when its service is unreachable; the full rubric-evolution loop is the production aim, not yet the production reality.

In the World →

Inputs

L1–L7 extraction · teacher rationales

Outputs

Symbolic reasoning chains · verifiable step graphs

Worked exampleAn L7 tacit-knowledge instance ('never hot-swap that sensor while the loop is armed') goes to the teacher, which returns a rationale; the Synthesizer splits it into premise (loop armed implies live current), rule (hot-swap under load arcs), constraint (vendor spec §4 voltage window), cross-reference (matches two incident reports in the corpus), conclusion (disarm first). Three such chains are sampled for the one instance and all three become train.jsonl rows.

In the literature

The vocabulary

kice-tice

KICE + TICE

Stage 02 — twin agents mine the corpus in parallel across seven layers of knowledge, from rare concepts down to tribal know-how.

in the dictionary →

sft-raft

SFT + RAFT

Stage 04: LoRA fine-tuning where every batch carries oracle documents plus deliberate distractors, so the student internalizes grounded reasoning instead of memorized answers.

in the dictionary →

scotd

SCoTD

Symbolic Chain-of-Thought Distillation: train a small student on many sampled teacher rationales, so models far below the CoT-emergence scale still learn to reason step-by-step.

in the dictionary →

autoresearch

AutoResearch

The meta-agent that evolves Stage 03's reasoning-step templates — and the extraction rubrics upstream — from failure analysis, cycle over cycle.

in the dictionary →

For agents

$ curl "https://qukaizen.com/what?term=scotd-cotd&world=nucleus"

# the raw compiled artifact: /worlds/nucleus/terms.json