MeTTa-NARS Full

help edit space_dashboard
Approved by Ursula Addison on 2026-05-14

Source Verdict

Per Non-clustered HAA cluster pilot Source 5 close 2026-05-07: [IMPLEMENTATION-BACKED-NAL1-5] / [METTA-NARS-NOT-PLN] / [MOTTO-POC-INTEROP].

MeTTa-NARS is patham9's MeTTa port of the Non-Axiomatic Reasoning System (NARS) tradition (Pei Wang) — specifically derived from the OpenNARS-for-Applications (ONA) control architecture. Implements NAL-1 through NAL-5 inference layers in MeTTa with frequency/confidence (f/c) two-dimensional truth values. Paradigm-distinct from PLN (closed cluster pilot 2026-04-25); the truth-value semantics, control structure, and inference primitives are NARS-faithful, not PLN-equivalent. Cross-link to PLN as paradigm boundary only.

Core Mechanisms

  • NAL-1..NAL-5 inference layers. Source files implement the canonical NARS truth functions (deduction, induction, abduction, comparison, analogy) and conversion rules, with the f/c truth-value pair carried through inference per NAL semantics.
  • ONA-derived control. Control structure mirrors OpenNARS-for-Applications (Hammer's NARS implementation) — selection of beliefs by priority/durability, attention buffer management, time-bounded reasoning under the Assumption of Insufficient Knowledge and Resources (AIKR).
  • MeTTa-native types. Truth values, beliefs, and inference rules expressed as MeTTa atoms; rules implemented as MeTTa rewrite patterns. Allows NARS to compose with other MeTTa code on the same runtime.
  • NARS-GPT POC interop. Per the parent +deep_dive sibling, MeTTa-NARS includes a NARS-GPT integration POC. Treat as proof-of-concept, not production behavior.

Implementation Surface

RepositoryHEADDateVerdict
patham9/metta-nars945e2f82024-10-14[IMPLEMENTATION-BACKED-NAL1-5] — primary maintained line in patham9's portfolio. README confirms ONA-style NARS implementation in MeTTa with non-axiomatic logic + multi-step reasoning examples.

Author: Patrick Hammer (patham9). MeTTa-NARS is part of patham9's "non-axiomatic" portfolio together with NACE and Hammer's co-authorship on the AIRIS paper (Cook & Hammer 2024). NACE has a bridge.py ONA bridge surface that interfaces with MeTTa-NARS / OpenNARS — same author, complementary projects.

Cluster-Narrative Position — Paradigm Boundaries

vs PLN (closed cluster pilot 2026-04-25)

NARS truth values are frequency/confidence pairs (f, c); PLN truth values are strength/count pairs (s, n). The semantics, evidence accumulation rules, inference operators, and control regimes are different. The Goertzel mattermost paper "Comparing Truth Value Formulas in PLN and NARS Under High Uncertainty of Node Probabilities" attempts a comparison framework, but per HAA Source 5: [PAPER-NOT-IMPLEMENTED-IN-METTA-NARS] — paper-cited equations / uncertainty-mapping formulas were not located in the metta-nars source at HEAD via cross-grep, AND the paper carries a hallucination warning (GPT-o1-produced). Treat any "PLN ≈ NARS" claim from that paper with skepticism.

Mirror of S1 V1-4 [AIRIS-CONFIDENCE-NOT-PLN-TV]: PLN-comparable reasoning frameworks should be treated as paradigm-distinct, not as implementations or successors of PLN.

vs AIRIS (closed HAA Source 1 2026-05-06)

Different paradigms despite shared author (Hammer co-authored AIRIS; Hammer authored MeTTa-NARS). AIRIS uses curiosity/state-graph/AIRIS-confidence apparatus per closed S1 verdicts (V1-3..V1-5); MeTTa-NARS implements NAL inference. The two are NOT collapsible.

vs NACE (HAA Source 5 2026-05-07)

Shared author (patham9), shared NAL truth-value semantics. Different roles: MeTTa-NARS is a faithful NAL-1..5 reasoner port; NACE is a causal-rule learning agent that uses NAL evidence values for tracking. NACE bridges TO MeTTa-NARS / ONA, but they are separate projects (not one wrapping the other). See NACE.

vs MeTTa-Motto (cross-org sweep candidate)

Per closed S2 reconciliation, MeTTa-Motto stays out of HAA scope and defers to cross-org sweeps. The metta-nars POC interop with Motto is a peripheral integration note, NOT a reason to fold Motto into HAA. (V5-14 narrative-guard at S5 close.)

What This Card Is Not

  • Not a PLN implementation, not a PLN successor, not a PLN-equivalent. NARS f/c TV is paradigm-distinct from PLN s/c TV. The two coexist in the wiki; cross-linking is at the paradigm-boundary level, not at the implementation level.
  • Not the same as NACE. NACE is a learning agent that uses NAL evidence; MeTTa-NARS is the reasoner. They are sibling projects in patham9's portfolio.
  • Not a complete NAL-1..NAL-9 implementation. NAL-1 through NAL-5 are implementation-backed at HEAD; higher NAL layers (NAL-6 through NAL-9) are not in scope at this version. Reviewers may verify higher-NAL coverage in future close work.
  • Not the canonical OpenNARS. MeTTa-NARS is patham9's MeTTa port derived from the ONA tradition. The C-side OpenNARS / OpenNARS-for-Applications repos are upstream-historical, not part of this MeTTa wiki coverage.

Status and Sources

Implementation status: Maintained research line. Latest commit "Update: filter timing indicators" (2024-10-14) suggests recent control-loop refinement. README confirms NAL non-axiomatic logic + multistep reasoning examples.

Primary sources:

  • Wang, P. (2013). Non-Axiomatic Logic: A Model of Intelligent Reasoning. World Scientific. Canonical NARS bibliography.
  • Hammer, P. (n.d.). MeTTa-NARS: README + repository at patham9/metta-nars.
  • Goertzel, B. (n.d.). Comparing Truth Value Formulas in PLN and NARS Under High Uncertainty of Node Probabilities. Mattermost paper. Treat as boundary-input only — paper carries GPT-o1 hallucination warning per project memory; cross-validate any extracted claim against this card + Pei Wang NARS bibliography + closed PLN cluster pilot before adopting.

Cluster-pilot extraction archive: scripts/archive/non_clustered_haa_pilot/source5_NACE_AIDSL_MOSES_MeTTaNARS/ (close 2026-05-07). V5-11 [METTA-NARS-IMPLEMENTATION-BACKED-NAL1-5] + V5-12 [METTA-NARS-NOT-PLN] + V5-13 [PATHAM9-HAA-PORTFOLIO] + V5-14 [METTA-MOTTO-OUT-OF-S5] carry-forwards locked at S5 close.