Research catalog
The 43 VolvenceZero research candidates imported from docs/moving forward/探索方向.md. Each candidate carries adoptable suggestions that can be turned into plugins + A/B runs.
| Code | Title | Group | Priority | R-IDs | Evaluation | Status | ROI |
|---|---|---|---|---|---|---|---|
| CMA-1 | derived layer 引入只读 spreading activation shadow path5 actions | CMA | P2/M-L | — | Spec-level | PROPOSED | — |
| CMA-2 | 实现 VZ-MemProbe 评估套件（CMA 4 探针 VZ 化）5 actions | CMA | P1/M | — | Spec-level | PROPOSED | — |
| CMA-3 | ModificationGate 加 memory drift telemetry / anomaly report4 actions | CMA | P1-P2/S-M | — | Spec-level | PROPOSED | — |
| COG-10 | Plasticity-stability 作为长期能力：巩固、遗忘边界、可逆可塑性3 actions | COG | P1-P2/M | — | Spec-level | PROPOSED | — |
| COG-11 | Agentic EQ / relationship 长程评估：证明关系存在质量，而不只是任务成功4 actions | COG | P1/M | — | Spec-level | PROPOSED | — |
| COG-4 | Environment Interface 闭环：Observe → Act → Assimilate → PE 作为 typed world feedback3 actions | COG | P1/M | — | Spec-level | PROPOSED | — |
| COG-5 | R13 压缩-强化交替作为一等工程程序，而非隐含原则3 actions | COG | P1/M | — | Spec-level | PROPOSED | — |
| COG-6 | Epistemic PE / Active Inference routing：区分可学习误差与噪声，再驱动 β_t / curiosity3 actions | COG | P1/M | — | Spec-level | PROPOSED | — |
| COG-7 | Latent reasoning / test-time compute budget：何时多想、在哪个 latent 层多想、何时停止3 actions | COG | P1/S-M | — | Spec-level | PROPOSED | — |
| COG-8 | Mechanistic interpretability 作为 owner health / drift readout，而非运行时任意 steering3 actions | COG | P1/M | — | Spec-level | PROPOSED | — |
| COG-9 | 发育式 / ontogenetic curriculum：单个 lifeform 的阶段性成熟与能力保持3 actions | COG | P2/M | — | Spec-level | PROPOSED | — |
| DM-1 | PE 输出从 scalar 升级为 distribution3 actions | DM | — | — | Spec-level | PROPOSED | — |
| DM-2 | z_t 接口采用 successor-feature behavior basis3 actions | DM | — | — | Spec-level | PROPOSED | — |
| DM-3 | regime trigger 用 interest function 替代 hard-coded3 actions | DM | — | — | Spec-level | PROPOSED | — |
| DM-4 | ModificationGate 必须 anchor 在 verifiable evaluation3 actions | DM | — | — | Spec-level | PROPOSED | — |
| DM-5 | background-slow 反思引入 imagination rollout4 actions | DM | — | — | Spec-level | PROPOSED | — |
| DM-6 | episodic→persistent 晋升用 meta-learned data value3 actions | DM | — | — | Spec-level | PROPOSED | — |
| DM-7 | evaluation 引入"代际比较"作为开放任务的进步度量3 actions | DM | — | — | Spec-level | PROPOSED | — |
| EVO-1 | ModificationGate 候选 archive = MAP-Elites + island（行为描述符 × elite + 岛间限速迁移）3 actions | EVO | P1/M | — | Spec-level | PROPOSED | — |
| EVO-2 | Evaluation cascade（廉价→昂贵）+ LLM-judge **仅 readout**（自然性/简洁性）3 actions | EVO | P0/M | — | Spec-level | PROPOSED | — |
| EVO-3 | Scenario package = QD behavior archive（regime × 强度 × rupture 格点 elite）3 actions | EVO | P1/M | — | Spec-level | PROPOSED | — |
| EVO-4 | Meta-prompt 在 expression 层 SHADOW 内演化（不改 z_t/β_t）3 actions | EVO | P2/S | — | Spec-level | PROPOSED | — |
| EVO-5 | Proposer / verifier 不对称共演化（PAIRED 式 regret 引导场景库；evaluator 只读）3 actions | EVO | P2/M | — | Spec-level | PROPOSED | — |
| EVO-6 | Cross-generation winrate / head-to-head 作为 ModificationGate 硬证据（与 DM-7 合并叙述）0 actions | EVO | P0/S | — | Spec-level | PROPOSED | — |
| OA-1 | spec motivation 引用 N3/N4/N6 实证3 actions | OA | P0/S | — | Spec-level | PROPOSED | — |
| OA-10 | character-soul-bootstrap 加 value-prioritization-as-regime4 actions | OA | P2/S | — | Spec-level | PROPOSED | — |
| OA-2 | 在 expression-layer spec 形式化 Mind/Face 隔离3 actions | OA | P0/S | — | Spec-level | PROPOSED | — |
| OA-3 | ModificationGate 加 framing-aware 检查（inoculation 工程化）6 actions | OA | P0/M | — | Spec-level | PROPOSED | — |
| OA-4 | 实现 VZ-Audit Agent 作为 ModificationGate 标配4 actions | OA | P1/M | — | Spec-level | PROPOSED | — |
| OA-5 | 实现 VZ-Spec-Stress 工具（cross-instance disagreement）5 actions | OA | P1/M | — | Spec-level | PROPOSED | — |
| OA-6 | 给 metacontroller 加 prompt-injection 抗操控测试5 actions | OA | P1/M | — | Spec-level | PROPOSED | — |
| OA-7 | 给 derived 层加 chunky-routing 自检（SURF）5 actions | OA | P1/M | — | Spec-level | PROPOSED | — |
| OA-8 | reflection engine 显式"用弱模型 + 工具"0 actions | OA | P2/S | — | Spec-level | PROPOSED | — |
| OA-9 | 探索 PE faithfulness 指标（N6 思想 → R-PE 工程化）5 actions | OA | P2/L | — | Spec-level | PROPOSED | — |
| SYS-2 | ModificationGate 引入 VC 容量上限 (Capacity Bound) 双门机制：防过拟合与泛化丧失3 actions | SYS | P0/M | — | Spec-level | PROPOSED | — |
| SYS-3 | KV Cache Steering 作为无梯度潜空间注入口：Mind 到 Face 的物理控制通道3 actions | SYS | P1/M | — | Spec-level | PROPOSED | — |
| SYS-4 | 引入 A-Mem 的 Zettelkasten 卡片演化 Schema：让 Derived Layer 有坚实的图谱结构3 actions | SYS | P1/M | — | Spec-level | PROPOSED | — |
| SYS-5 | Latent Action RL 作为控制器强化基线：在冻结 LLM 上挂极小控制器跑 RL3 actions | SYS | P0/L | — | Spec-level | PROPOSED | — |
| SYS-6 | 躯体标记 (Somatic Markers) 与全局工作区 (GNWT) 竞争机制：注意力竞争与生理偏置3 actions | SYS | P2/L | — | Spec-level | PROPOSED | — |