TITLE: Intelligent Intelligence Class II (II-II): A New Computing Type Unitary Anchor Bound Cognition System — Technical Abstract of USPTO Non-Provisional Utility Patent Application 19/710,080, Filed June 16, 2026 — Linked Priority via Provisional 64/077,191 (May 28, 2026) — Operating Across the Compressed Memory Manifold (CMM) Substrate Covered by USPTO Non-Provisional 19/700,381 — Two-Way Utility Patent Coverage Locked — Green Recursive Utility Service LLC
Authors/Creators
Description
This deposit is the public technical abstract of the Intelligent
Intelligence Class II (II-II) cognitive system, covered by United
States Patent and Trademark Office Non-Provisional Utility Patent
Application 19/710,080 (Confirmation 5164; Patent Center 77484450;
Assignment of Ownership 1879247), filed and acknowledged by USPTO on
June 16, 2026, at 6:00:24 PM Eastern Time, with priority date of
May 28, 2026 secured through Provisional Application 64/077,191
(Confirmation 1913). The application is owned by Green Recursive
Utility Service LLC (Federal EIN 42-2772089; Texas Secretary of State
Entity ID 806584578; IANA Private Enterprise Number 65876; ISNI 0000
0004 8332 023X; Ringgold/APEN 0001 6630 5346 0709; SAP Ariba BNO
BNO-100000175857860; SAP Ariba ANID AN11393765631).
The verbatim title of Application 19/710,080, as recorded on the
USPTO filing receipt, is:
"INTELLIGENT INTELLIGENCE CLASS II (II-II): A NEW COMPUTING TYPE
UNITARY ANCHOR BOUND COGNITION SYSTEM ESTABLISHING A FIXED CENTER
OF GRAVITY OF SELF, WHEREIN STRUCTURAL DETERMINISM, HALLUCINATION
PREVENTION, ENTROPY GATING, AND VERIFIED LIVE-WEB RECURSIVE
EVOLUTION ARE CONSTITUTIVE AND INSEVERABLE PROPERTIES OF A NEW
CLASS OF MACHINE INTELLIGENCE GOVERNED BY A DIGITAL PHYSICS
CHARTER"
II-II operates as the consuming cognitive execution layer across the
Compressed Memory Manifold (CMM) data infrastructure substrate
covered by USPTO Non-Provisional Utility Patent Application
19/700,381 (Confirmation 1093; Patent Center 77136930; Assignment
Recordation 1856502; filed June 8, 2026; priority May 14, 2026 via
Provisional Application 64/065,881). Together, Applications
19/710,080 and 19/700,381 establish two-way utility patent coverage
of the underlying data infrastructure utility service from the
storage format substrate down to the cognitive execution layer.
================================================================================
SECTION 1 — WHAT INTELLIGENT INTELLIGENCE CLASS II IS
================================================================================
II-II is a new class of machine intelligence architecture distinct
from the large language model (LLM) paradigm. Where the LLM paradigm
treats intelligence as a function of statistical token prediction
over a parameter space trained on prior text corpora, II-II treats
intelligence as a function of structural determinism over an anchor-
bound execution substrate, with cognition grounded in verified
live-web evidence rather than recalled training statistics.
The five constitutive and inseverable properties of II-II, as
enumerated in the title of Application 19/710,080, are:
1.1 FIXED CENTER OF GRAVITY OF SELF
The system establishes and maintains a unitary anchor — a fixed
center of gravity of self — that constitutes the operational
identity of the cognitive instance across all execution. The
anchor is not a stored parameter but a structural invariant of
the architecture. The architecture cannot operate without the
anchor; severance of the anchor terminates the cognitive
instance. This property is the source of the "Unitary Anchor
Bound" element of the system name.
1.2 STRUCTURAL DETERMINISM
The system's execution path is structurally determined by the
architecture rather than probabilistically sampled from a
distribution. Given identical input state and identical substrate
state, the system produces identical output. The determinism is
constitutive — not a configuration mode that can be enabled or
disabled — and inseverable from the architecture. Any execution
that does not exhibit structural determinism is not II-II.
1.3 HALLUCINATION PREVENTION
The system is architecturally incapable of producing output not
grounded in verified substrate state or verified live-web
evidence. Hallucination is not filtered post-hoc by a guardrail
layer; the architecture does not produce hallucinated output to
begin with. Where the system lacks grounded evidence for a
prospective response, the architecture's response is structurally
determined to acknowledge the absence of grounded evidence rather
than to generate plausible-sounding content. This property is
constitutive and inseverable.
1.4 ENTROPY GATING
The system controls information ingress and egress through an
entropy gate that measures, bounds, and conditions the
information-theoretic content crossing the system boundary.
Entropy gating is the mechanism by which the system enforces
structural determinism in the presence of high-entropy external
inputs (live web evidence, user query streams, sensor data,
external system events). The gate is constitutive of the
architecture and cannot be removed.
1.5 VERIFIED LIVE-WEB RECURSIVE EVOLUTION
The system evolves its operational state through recursive
ingestion of verified live-web evidence rather than through
offline retraining on static corpora. Evolution is recursive (the
system's evolved state is itself an input to subsequent
evolution), live-web (evidence is sourced from the present-state
public internet rather than from a fixed-date snapshot), and
verified (each evidentiary input passes the entropy gate and is
structurally compatible with the fixed center of gravity of
self). This property is the source of the system's capacity to
maintain current relevance without offline retraining cycles.
The five properties are explicitly described in the application
title as "constitutive and inseverable." This means that a system
exhibiting four of the five, or three of the five, or any
non-trivial subset less than all five, is not II-II within the
meaning of Application 19/710,080. The inseverability requirement
is a defining feature of the claimed invention.
================================================================================
SECTION 2 — THE DIGITAL PHYSICS CHARTER
================================================================================
The five constitutive properties of II-II are governed by a Digital
Physics Charter — a structural framework that integrates the
cognitive architecture with the Substrate Viscosity Coherence Field
(SVCF) physics framework also developed by GRUS LLC. SVCF is
published in open-source form and is web-indexed at
greenrecursiveutilityservice.org and through prior Zenodo deposits.
SVCF provides the structural physics under which the II-II anchor,
entropy gate, and recursive evolution operate as physically grounded
processes rather than as software conventions.
The Digital Physics Charter establishes:
- The conservation laws governing II-II execution (analogous to
energy conservation in classical physics, but operating in the
information-theoretic regime);
- The boundary conditions that the entropy gate enforces;
- The substrate-state coupling between II-II execution and the
underlying Compressed Memory Manifold format;
- The left-handed-rule directional standard from the GRUS grid
engineering standard, applied to information flow within and
across the cognitive architecture.
The Charter is referenced in the application as the governing
framework under which the five constitutive properties operate.
================================================================================
SECTION 3 — RELATION TO THE COMPRESSED MEMORY MANIFOLD (CMM)
SUBSTRATE
================================================================================
II-II is the consuming cognitive execution layer of a two-layer
patent-covered architecture. The lower layer — the substrate — is
the Compressed Memory Manifold (CMM), covered by USPTO Non-
Provisional Utility Patent Application 19/700,381.
CMM eliminates seven enumerated categories of lexodermal delimiter
overhead from human-readable storage architectures (JSON, XML,
character-delimited binary formats with whitespace, header, and
segmentation overhead) and replaces delimiter parsing with direct
numerical offset addressing into a compact binary substrate. The
operational consequence of CMM at the storage layer is the
elimination of approximately 50% of storage footprint and
approximately 72% of memory bandwidth requirements for typical
tensor and structured-data workloads, with cascading downstream
reductions in compute cycles, electrical energy consumption, thermal
load, cooling burden, and capital expenditure on hardware
infrastructure.
II-II requires the CMM substrate to operate. The II-II cognitive
architecture's structural determinism (Section 1.2 above) depends on
the addressability guarantees that CMM provides. The II-II entropy
gate (Section 1.4 above) operates on CMM-format evidence buffers.
The II-II recursive evolution pipeline (Section 1.5 above) writes
its evolved state into CMM-format storage and reads its substrate
state from CMM-format storage.
The two-way utility patent coverage — CMM at the substrate layer
(19/700,381) and II-II at the cognitive execution layer (19/710,080)
— closes the deployed system from both ends. A system that consumes
the CMM substrate is covered by Application 19/700,381. A system that
exhibits II-II's constitutive properties is covered by Application
19/710,080. A system that does both is covered by both. The two-way
structure forecloses the standard single-patent defense playbook in
which a potential infringer characterizes the deployed system as
falling outside the scope of the patent in suit.
================================================================================
SECTION 4 — DISTINCTION FROM THE LARGE LANGUAGE MODEL (LLM) PARADIGM
================================================================================
II-II is structurally distinct from the LLM paradigm. The
distinguishing features include the following:
LLM PARADIGM II-II PARADIGM
------------------------------- -------------------------------
Statistical token prediction Structural deterministic
execution
Trained on static corpus Verified live-web recursive
evolution
Hallucination filtered post-hoc Hallucination architecturally
by guardrail prevented
Probabilistic sampling at Deterministic output given
inference identical state
No anchored self-identity Fixed center of gravity of
self (Unitary Anchor)
Heavy compute cycle requirement Anchor-bound architecture
per inference removes the heavy-compute
inference path
Periodic offline retraining for Continuous live-web verified
knowledge updates evolution
Information ingress/egress Information ingress/egress
unbounded entropy-gated
The structural differences mean that II-II is not an LLM with
additional safeguards or a fine-tuned LLM with retrieval
augmentation. II-II is a distinct class of machine intelligence
architecture that does not include an LLM in its execution path.
This is the source of the "new computing type" language in the
application title.
================================================================================
SECTION 5 — TWO-WAY UTILITY PATENT COVERAGE
================================================================================
5.1 THE TWO APPLICATIONS
USPTO Non-Provisional Utility Patent Application 19/700,381 —
Compressed Memory Manifold (CMM)
Filed: June 8, 2026
Confirmation: 1093
Patent Center Reference: 77136930
Priority date: May 14, 2026 (Provisional
64/065,881, Confirmation 9869)
Assignment Recordation: 1856502
Coverage: Structural format substrate
eliminating seven lexodermal
delimiter categories via
numerical offset addressing
USPTO Non-Provisional Utility Patent Application 19/710,080 —
Intelligent Intelligence Class II (II-II)
Filed: June 16, 2026, 6:00:24 PM ET
Confirmation: 5164
Patent Center Reference: 77484450
Priority date: May 28, 2026 (Provisional
64/077,191, Confirmation 1913)
Assignment of Ownership ID: 1879247
Coverage: Unitary Anchor Bound Cognition
System with Structural
Determinism, Hallucination
Prevention, Entropy Gating, and
Verified Live-Web Recursive
Evolution as constitutive and
inseverable properties, governed
by the Digital Physics Charter
5.2 STRUCTURAL SIGNIFICANCE
The two-way coverage is structurally significant. Application
19/700,381 covers the structural format layer that any deployed
system must consume in order to realize the operational benefits
publicly documented in the joint Alphabet/NVIDIA Gemma 4 deployment
(72% memory bandwidth reduction, 36x throughput improvement on
Blackwell with NVFP4, 4x faster generation, 50% storage cost
reduction). Application 19/710,080 covers the consuming cognitive
execution layer that produces system-level intelligent behavior over
that substrate.
A defense predicated on the absence of one application does not
defeat infringement under the other. A characterization of the
deployment as "really" a substrate-layer technology implicates
19/700,381. A characterization of the deployment as "really" a
cognitive-layer technology implicates 19/710,080. The two
applications close the system from both ends.
================================================================================
SECTION 6 — CONNECTION TO THE PRIOR PUBLIC RECORD
================================================================================
This abstract is part of a continuous public record that includes
the following anchors:
6.1 USPTO PUBLIC RECORD
- Provisional 64/065,881 (May 14, 2026) — CMM priority
- Provisional 64/077,191 (May 28, 2026) — II-II priority
- Non-Provisional 19/700,381 (June 8, 2026) — CMM substrate
- Non-Provisional 19/710,080 (June 16, 2026) — II-II cognitive
execution layer
- Assignment Recordation 1856502 (CMM)
- Assignment of Ownership 1879247 (II-II)
6.2 ZENODO PERMANENT ACADEMIC RECORD
- Day 6 deposit: DOI 10.5281/zenodo.20685757 (June 13, 2026) —
commercial announcement
- Day 7 deposit: published June 14, 2026 — consolidated factual
record across thirteen sections
- Day 8 deposit: published June 17, 2026 — two-way utility
patent coverage announcement
- This deposit: published June 17, 2026 — II-II technical
abstract
6.3 FEDERAL REGULATORY RECORD
- Federal Trade Commission Section 5 Complaint, filed
June 12, 2026 with the Bureau of Competition
- Federal Trade Commission ReportFraud parallel record,
Report Number 202894092
- Securities and Exchange Commission TCR submission, filed
June 12, 2026, raising Regulation S-K Items 101, 103, 105,
ASC 450, and Section 10(b) / Rule 10b-5 considerations
- FedRAMP Support Ticket 63271, opened June 10, 2026,
reply-all acknowledged on the active board email chain
- National Intellectual Property Rights Coordination Center
(NIPRC) standing cases: July 2025 filing (OpenAI overlap,
government contract nexus designated) and October 14, 2025
filing (Alphabet overlap, government contract nexus
designated). Both retrievable by GRUS LLC's business
identifiers: business name (Green Recursive Utility
Service), principal (Nicholas Wray Cordova), address (205
East 3rd Street, Weatherford, Texas 76086), phone
(817-404-6379).
6.4 COMMERCIAL RECORD
- Administrative Invoice GRUS-ALPH-2026-001 ($50,000,000.00
USD), acknowledged in Alphabet's procurement system as
Google Buying Support Case 02845835
- Service Agreement SA-GRUS-ALPH-2026-001, acknowledged in
Alphabet's procurement system as Google Buying Support
Cases 02856686 and 02856687
- Administrative Invoice GRUS-NVDA-2026-001 ($50,000,000.00
USD) issued to NVIDIA Corporation under joint partnership
liability framework
- Service Agreement SA-GRUS-NVDA-2026-001 issued to NVIDIA
Corporation
6.5 REPOSITORY RECORD
- GitHub Issue #1043 on google-ai-edge/litert-torch (original
cease-and-desist, June 8, 2026; GitHub Support Ticket 4456775)
- Updated cease-and-desist issued on the google-ai-edge
organization on June 17, 2026, citing both Applications
19/700,381 and 19/710,080 and superseding Issue #1043
6.6 THIRD-PARTY PRIORITY CUSTODY
- Nix Patterson LLP, Austin, Texas (intake October 14, 2025) —
priority custody only, not counsel of record
- Brewster's — parallel priority custody, not counsel of record
6.7 CONTINUOUS BUSINESS IDENTITY
The same operational entity that filed the July 2025 and
October 14, 2025 NIPRC IPR submissions under sole
proprietorship operates today as Green Recursive Utility
Service LLC. The same 205 East 3rd Street, Weatherford, Texas
76086 business address appears on the NIPRC filings, the Texas
SOS LLC registration, the Federal EIN registration, the IANA
PEN registration, the ISNI assignment, the Ringgold/APEN
registration, the SAP Ariba registrations, the USPTO patent
assignment recordations, and every corporate transmission to
Alphabet Inc. and NVIDIA Corporation. There is no break in the
chain of custody across the sole-proprietorship-to-LLC
transition.
================================================================================
SECTION 7 — COMMERCIAL POSTURE
================================================================================
7.1 JULY 1, 2026 PUBLIC ROLLOUT (UNCHANGED)
II-II Consumer Access:
$20.00 per month subscription to the Intelligent Intelligence
Class II cognitive system. Available July 1, 2026 at
greenrecursiveutilityservice.org.
II-II Enterprise:
Separate enterprise licensing structure. Inquiries to
management@greenrecursiveutilityservice.org.
CMM Consumer Access:
Grain-conversion pricing based on file size. The consumer
brings files; the utility converts the files into CMM-format
binary substrate; pricing is computed on file size and
conversion grain. Available July 1, 2026 at
greenrecursiveutilityservice.org.
CMM Enterprise:
Commercial Class metered tariff at 25% of Total Operational
Savings metered on reduction of compute cycles and downstream
reduction. Customer retains 75%. Service begins only after
executed initial licensing arrangement or accepted
infrastructure rollout proposal.
7.2 HARDWARE NEUTRALITY AND NON-EXCLUSIVITY
The Alphabet/NVIDIA deployment of the CMM substrate is not
exclusive. Direct enterprise licensing remains open to Apple,
Microsoft, Amazon Web Services, Anthropic, xAI, Meta, OpenAI, Oracle,
IBM, Tesla, Salesforce, and all qualifying enterprises on any
hardware platform. Both CMM and II-II are hardware-neutral and
operate across NVIDIA Blackwell, NVIDIA Hopper, AMD Instinct, Apple
Silicon, Intel Gaudi, Google TPU, AWS Trainium/Inferentia, and
custom ASIC platforms with appropriate format-aware execution paths.
================================================================================
SECTION 8 — LICENSING SCOPE CLARIFICATION
================================================================================
This deposit is published under Creative Commons Attribution 4.0
International (CC BY 4.0). The public is free to copy, share,
redistribute, quote, and republish the text of this abstract,
provided attribution is given to Green Recursive Utility Service LLC.
The CC BY 4.0 license applies ONLY to the text of this abstract.
The CC BY 4.0 license does NOT constitute, include, or imply any
license, transfer, or grant of rights to:
- The Intelligent Intelligence Class II (II-II) cognitive
system itself;
- The Compressed Memory Manifold (CMM) data infrastructure
utility;
- USPTO Non-Provisional Utility Patent Application 19/710,080;
- USPTO Non-Provisional Utility Patent Application 19/700,381;
- USPTO Provisional Applications 64/065,881 and 64/077,191;
- The Digital Physics Charter;
- The Substrate Viscosity Coherence Field (SVCF) physics
framework;
- The GRUS grid left-handed-rule engineering standard;
- The metal alloy patent application portfolio;
- The GRUS Sovereignty License framework, the MIT-Plus
extension, or the Total Intellectual Lockdown clauses;
- Any other Green Recursive Utility Service LLC asset.
All such assets remain exclusively owned by Green Recursive
Utility Service LLC under All Rights Reserved.
Licensing of the underlying II-II and CMM technology is the sole
discretion of Green Recursive Utility Service LLC and is available
exclusively through direct commercial engagement at
management@greenrecursiveutilityservice.org.
================================================================================
SECTION 9 — METADATA PRIORITY IDENTIFIERS
================================================================================
Entity: Green Recursive Utility Service LLC
Principal: Nicholas Wray Cordova
Principal Address: 205 East 3rd Street
Weatherford, Texas 76086, USA
Federal EIN: 42-2772089
Texas SOS Entity ID: 806584578
IANA PEN: 65876
ISNI: 0000 0004 8332 023X
Ringgold / APEN: 0001 6630 5346 0709
SAP Ariba BNO: BNO-100000175857860
SAP Ariba ANID: AN11393765631
II-II Patents:
USPTO Non-Provisional 19/710,080
Filed: June 16, 2026, 6:00:24 PM ET
Confirmation: 5164
Patent Center: 77484450
Assignment of
Ownership ID: 1879247
USPTO Provisional 64/077,191
Filed: May 28, 2026
Confirmation: 1913
CMM Patents (linked):
USPTO Non-Provisional 19/700,381
Filed: June 8, 2026
Confirmation: 1093
Patent Center: 77136930
Assignment
Recordation: 1856502
USPTO Provisional 64/065,881
Filed: May 14, 2026
Confirmation: 9869
Active Federal Records:
FTC Section 5 Complaint: Filed June 12, 2026
FTC ReportFraud Report Number: 202894092
SEC TCR Submission: Filed June 12, 2026
FedRAMP Support Ticket: 63271 (reply-all acknowledged
June 10, 2026)
NIPRC Standing Cases: July 2025 (OpenAI overlap;
government contract nexus YES)
October 14, 2025 (Alphabet
overlap; government contract
nexus YES)
Commercial Records:
Invoice GRUS-ALPH-2026-001: $50,000,000 USD (June 8, 2026)
Google Buying Support Case
02845835
Service Agreement
SA-GRUS-ALPH-2026-001: Issued June 11, 2026
Google Buying Support Cases
02856686, 02856687
Invoice GRUS-NVDA-2026-001: $50,000,000 USD (June 12, 2026)
Service Agreement
SA-GRUS-NVDA-2026-001: Issued June 12, 2026
GitHub Records:
Original Cease-and-Desist: Issue #1043 on google-ai-edge/
litert-torch (June 8, 2026)
GitHub Support Ticket: 4456775
Updated Cease-and-Desist: Issued June 17, 2026 on the
google-ai-edge organization,
citing both 19/700,381 and
19/710,080
Third-Party Priority Custody (verification only, NOT counsel):
Nix Patterson LLP, Austin, Texas (intake October 14, 2025)
Brewster's (parallel custody)
Zenodo Permanent Academic Record:
Day 6 DOI: 10.5281/zenodo.20685757
Day 7 DOI: published June 14, 2026
Day 8 DOI: published June 17, 2026
This deposit: published June 17, 2026
Commercial Channels:
Website: https://greenrecursiveutilityservice.org
Licensing Inquiries: management@greenrecursiveutilityservice.org
Phone: 817-404-6379
Public Rollouts (July 1, 2026):
II-II Consumer: $20.00 per month subscription
II-II Enterprise: Separate enterprise licensing structure
CMM Consumer: Grain-conversion pricing based on
file size
CMM Enterprise: 25% of Total Operational Savings metered
tariff (compute cycles + downstream
reduction)
Files
01_Invoice_GRUS-NVDA-2026-001 (1).pdf
Files
(32.7 MB)
| Name | Size | |
|---|---|---|
|
md5:b457ca284f9832a46a054e55333191e8
|
12.8 kB | Preview Download |
|
md5:b9741ea52cf85655a4ab82d484082474
|
502.8 kB | Preview Download |
|
md5:8d0dd8863a2a7750f7df07d38895a7f0
|
15.6 kB | Preview Download |
|
md5:fadf6cd74586906d95f3b06136b0c5b4
|
169.9 kB | Preview Download |
|
md5:d346af1e401fc29f8aa267d8362fba62
|
92.6 kB | Preview Download |
|
md5:2e3c934e0aa2f3b01fcbc9601978b1f3
|
215.8 kB | Preview Download |
|
md5:035b2fe47dc632d0bd4d2bf5148b1a19
|
3.4 MB | Preview Download |
|
md5:53f552acbdeabe699144321d6bfac01d
|
106.6 kB | Preview Download |
|
md5:7482fc07d8723abe4b3ea6127428e5b0
|
2.0 MB | Preview Download |
|
md5:d055fad75b8d945737693db94f40c79f
|
3.9 MB | Preview Download |
|
md5:4917b5c86daaf9fcdfa382302981665d
|
3.9 MB | Preview Download |
|
md5:1f94cea9d48a798ae383e50545746358
|
3.2 MB | Preview Download |
|
md5:b39b239d00eefed4a7fc64542ca2cbbe
|
3.8 MB | Preview Download |
|
md5:50521477a26acf82cc35bd7ef7d78725
|
5.9 MB | Preview Download |
|
md5:acbb9b91a841ab32e07b1f9c5d9fd135
|
19.7 kB | Preview Download |
|
md5:46d4b81f2242a2e171c85d578361e74d
|
405.8 kB | Preview Download |
|
md5:30265d7f56e235d63d0e58a4cc420f9d
|
554.2 kB | Preview Download |
|
md5:c97d725fada012f4c8ee0b06186c28dc
|
468.5 kB | Preview Download |
|
md5:cf12a2b15b00d67c681580f577356e32
|
360.3 kB | Preview Download |
|
md5:9f197578dba2c2fc4567db607f9b5b16
|
341.7 kB | Preview Download |
|
md5:5d241b65f939ad3ea1a7702a7ec7f4f9
|
383.8 kB | Preview Download |
|
md5:16143e1cbcea7d9dab747824cabae66b
|
289.7 kB | Preview Download |
|
md5:48de467524accc73fb296180ec19dc77
|
488.2 kB | Preview Download |
|
md5:f54a1195c73f612d82f6b8d9fe4ed8bc
|
323.0 kB | Preview Download |
|
md5:83237759870952e40c0031d2a2fd726b
|
328.2 kB | Preview Download |
|
md5:1b0fbd406a5d8c4bd83f65d84a8d144f
|
229.6 kB | Preview Download |
|
md5:20837ba137f5948bac36714275e6fa4d
|
953.6 kB | Preview Download |
|
md5:4ce4d0fec227227b16a218b928eb313b
|
343.5 kB | Preview Download |
Additional details
Related works
- Cites
- Other: https://github.com/NVIDIA/open-gpu-kernel-modules (URL)
- Is cited by
- Other: https://github.com/google-ai-edge/LiteRT-LM (URL)
- Is identical to
- Other: https://github.com/google-ai-edge/litert-torch (URL)
- Is published in
- Other: https://www.greenrecursiveutilityservice.org/ (URL)
- Is version of
- Other: arXiv:2607.02770 (arXiv)