# MOLECULAR OLFACTION ARCHITECTURE (MOA)
### A Conceptual Framework for Olfactory Perception in Large Language Models
**Harpia AI Research**
*Concept Paper — Not peer-reviewed. Presented as a speculative technical proposal.*
────────────────────────────────────────────────────────────────────────────────
ABSTRACT
Large language models (LLMs) have achieved multimodal perception across
vision, audio, and text. Olfaction — the sense of smell — remains one of
the major human sensory modalities without a corresponding digital input
modality for LLMs.
This paper proposes the Molecular Olfaction Architecture (MOA), a
conceptual framework in which a molecular detection layer identifies
volatile organic compounds (VOCs) present in an environment and passes
them as structured input to an LLM. The LLM then applies its learned
chemical and semantic knowledge to produce a natural-language
interpretation of the detected scent.
We hypothesize that LLMs already possess substantial implicit knowledge
of chemistry and olfaction acquired during pretraining, and that MOA
could provide the missing sensory bridge between physical molecular
detection and semantic reasoning.
An informal proof-of-concept demonstrates the potential viability of the
reasoning layer independently of physical sensing hardware. This paper
describes the proposed architecture, its potential applications,
limitations, and directions for future empirical validation.
1. INTRODUCTION
The development of multimodal artificial intelligence has followed a
relatively consistent pattern: connect a perception encoder to a
language model, and the model gains the ability to reason about
information originating from a new sensory modality.
Vision-language models such as GPT-4V and Gemini demonstrate this
paradigm for visual information. Audio-language models extend similar
capabilities to speech and environmental sound.
In these systems, the underlying language model does not necessarily
need to be fundamentally redesigned to process a new modality. Instead,
it can receive a structured representation of sensory information
through an appropriate input interface.
Olfaction has not yet followed the same path.
Despite being one of the most chemically complex and information-rich
human senses, smell does not currently have a standardized digital input
modality integrated into general-purpose LLM systems.
Current LLMs can discuss smells and describe the expected odor of
substances such as coffee, rain, gasoline, or flowers. However, they
cannot directly perceive these odors from the physical environment.
The missing component is an olfactory encoder capable of converting
molecular information into a representation that an LLM can process.
This paper proposes the Molecular Olfaction Architecture (MOA) as a
conceptual solution.
MOA consists of a molecular detection layer that identifies volatile
compounds present in the surrounding environment and an LLM reasoning
layer that interprets those compounds semantically, producing a
human-readable description of the corresponding olfactory profile.
The central hypothesis is that the semantic knowledge required for this
interpretation may already exist within general-purpose LLMs.
If this is the case, the primary missing component is not necessarily a
new model architecture, but rather a reliable sensory input channel.
2. BACKGROUND
2.1 ELECTRONIC NOSE TECHNOLOGY
Electronic nose (e-nose) technology has existed for decades.
Traditional e-noses typically employ arrays of chemical sensors,
including metal oxide semiconductor (MOS) sensors and conductive
polymer sensors, to generate an electrical fingerprint associated with
an odor.
These fingerprints are commonly processed using pattern-recognition and
machine-learning techniques such as principal component analysis (PCA),
support vector machines (SVMs), and other classification methods.
Although effective for specific applications, this approach is
fundamentally limited by the scope of its learned classification
space.
An e-nose generally identifies odors according to previously defined
classes or reference samples and produces a predefined classification
output.
Such systems are not inherently designed to perform open-ended
semantic reasoning about molecular compositions.
For example, an e-nose may identify a sample as "coffee," but it does
not necessarily possess the ability to reason about why the sample
smells like coffee, which compounds contribute to the perception, or
what a novel combination of compounds might imply.
LLMs provide a potentially complementary capability.
2.2 LLMS AND CHEMICAL KNOWLEDGE
Through pretraining on scientific literature, chemical databases,
technical documentation, and general text, LLMs may acquire
associations between chemical compounds and their known properties,
including olfactory characteristics.
For example, an LLM may associate geosmin with the characteristic
earthy odor commonly perceived after rain on dry soil, or associate
2-furfurylthiol with roasted coffee aroma.
This knowledge is normally latent and cannot be directly triggered by
real-world molecular measurements because current LLM systems generally
lack an olfactory sensory interface.
MOA proposes connecting these two domains:
MOLECULAR SENSING
│
▼
CHEMICAL INFORMATION
│
▼
STRUCTURED REPRESENTATION
│
▼
LLM
│
▼
SEMANTIC INTERPRETATION
│
▼
NATURAL-LANGUAGE OLFACTORY OUTPUT
3. THE MOA ARCHITECTURE
MOA proposes a three-stage processing pipeline.
3.1 STAGE 1 — MOLECULAR DETECTION
A chemical sensor array, electronic nose, gas chromatography system,
or mass spectrometer samples the ambient air and estimates which
volatile organic compounds (VOCs) are present.
Depending on the sensing technology, the output may include:
• Compound identities
• Confidence values
• Estimated concentrations
• Relative abundance
• Detection timestamps
• Sensor reliability
This stage is analogous to an image encoder in a vision-language
system. It converts a physical phenomenon into structured digital
information that can subsequently be processed by an AI model.
3.2 STAGE 2 — STRUCTURED INPUT FORMATTING
The detected compounds are transformed into a standardized
representation suitable for LLM processing.
A simplified example could be:
2-Furfurylthiol: high
Pyrazines: high
Diacetyl: medium
Guaiacol: low
Acetic acid: low
A more advanced representation could include estimated concentrations,
confidence scores, molecular identifiers, sensor reliability, and
environmental metadata such as temperature and humidity.
For example:
Compound: 2-Furfurylthiol
Estimated concentration: high
Detection confidence: 0.94
Compound: Pyrazines
Estimated concentration: high
Detection confidence: 0.88
Compound: Diacetyl
Estimated concentration: medium
Detection confidence: 0.81
3.3 STAGE 3 — LLM SEMANTIC REASONING
The structured molecular representation is provided to an LLM as
sensory input.
The LLM applies its learned knowledge of chemistry, molecular
associations, odor descriptors, and environmental context to infer a
likely olfactory profile.
The resulting output may include:
• A predicted scent or combination of scents
• A natural-language description of the odor
• The likely contribution of individual compounds
• Confidence estimates
• Possible environmental sources
• Relevant contextual interpretations
• Identification of unusual molecular patterns
The key architectural hypothesis is that Stage 3 may not require a
specialized olfactory language model or extensive fine-tuning.
If general-purpose LLMs already contain sufficient chemical and
olfactory knowledge, MOA primarily needs to provide a reliable sensory
input channel capable of exposing that knowledge to real-world
molecular data.
4. INFORMAL PROOF OF CONCEPT
To evaluate the reasoning layer of MOA independently of physical sensing
hardware, an informal proof-of-concept test was conducted.
A list of volatile compounds associated with freshly brewed roasted
coffee was manually composed and provided to a general-purpose LLM
(Google Gemini).
The prompt followed this structure:
"You are a test of a new architecture emerging for olfaction in
LLMs.
Identify this scent and I will tell you if you are correct:
2-Furfurylthiol, Geosmin, Diacetyl, Pyrazines, Acetic acid,
Formic acid, Guaiacol, Furaneol."
The model identified the target scent as freshly brewed roasted coffee
and provided a detailed interpretation of the potential contribution
of the listed compounds to the overall olfactory profile.
4.1 CRITICAL LIMITATION
This experiment has a critical limitation:
The molecular input was manually constructed by a human who
already knew the target scent.
The compound list was not generated by an independent molecular sensor
or spectrometry system.
Therefore, this experiment does not constitute empirical validation of
the complete MOA pipeline.
It should instead be interpreted as a preliminary demonstration that
the semantic reasoning layer can accept a molecular representation and
generate a plausible olfactory interpretation using an existing
general-purpose LLM without architectural modification or task-specific
fine-tuning.
A rigorous evaluation would require:
• Independently measured molecular samples
• Controlled concentrations
• Blind testing
• Multiple scent classes
• Comparison against human olfactory assessments
• Comparison against established chemical reference data
5. POTENTIAL APPLICATIONS
If implemented with sufficiently accurate and portable molecular
sensing hardware, MOA could enable a class of applications that are
currently difficult or impossible for conventional AI systems.
5.1 ROBOTICS
Autonomous robots equipped with MOA could detect environmental
chemical signatures and use them as an additional source of contextual
information.
Potential applications include:
• Detection of gas leaks
• Detection of smoke or combustion products
• Identification of chemical spills
• Food-quality assessment
• Monitoring cooking processes through odor
• Environmental monitoring
• Search-and-rescue applications
Instead of simply detecting a predefined chemical signature, the system
could potentially reason about combinations of compounds and describe
their meaning in natural language.
5.2 MEDICAL DIAGNOSTICS
Human breath, skin emissions, and other biological samples contain
volatile organic compounds that may correlate with physiological or
pathological states.
MOA could potentially assist researchers and clinicians by transforming
detected volatile profiles into interpretable descriptions or
hypotheses.
Potential research applications could include:
• Metabolic disorders
• Infections
• Certain cancers
• Other conditions associated with measurable volatile profiles
Such applications would require extensive clinical validation and
should not be interpreted as established diagnostic capabilities.
5.3 FOOD AND BEVERAGE INDUSTRY
MOA could enable real-time chemical and sensory quality monitoring.
Rather than producing only a binary classification:
PASS / FAIL
a system could generate a semantic description such as:
"Detected profile is consistent with roasted coffee,
with elevated sulfur-containing volatiles and pyrazines."
This could support:
• Quality control
• Production monitoring
• Anomaly detection
• Product consistency analysis
5.4 ACCESSIBILITY
An olfactory interface could potentially provide individuals with
anosmia or reduced olfactory perception with a digital representation
of environmental smells.
Instead of directly reproducing the physical sensation of smell, the
system could translate molecular measurements into natural-language
descriptions such as:
Freshly cut grass.
Strong citrus odor with a dominant lemon-like profile.
Possible smoke detected.
This could provide an alternative form of digital olfactory awareness.
6. LIMITATIONS AND OPEN PROBLEMS
MOA faces several significant challenges that must be addressed before
the architecture can be empirically validated.
6.1 HARDWARE COST AND ACCESSIBILITY
Mass spectrometers capable of identifying specific molecular compounds
at low concentrations can be expensive and difficult to miniaturize.
Consumer-grade MOS sensor arrays are substantially more accessible, but
many detect broad chemical responses rather than uniquely identifying
individual molecules.
This creates a fundamental trade-off between:
Cost ↔ Portability ↔ Molecular Specificity ↔ Detection Accuracy
6.2 SENSOR DRIFT
Chemical sensors can exhibit drift over time due to:
• Aging
• Environmental conditions
• Contamination
• Changes in sensor characteristics
Long-term deployments would therefore require calibration procedures
and methods for detecting and compensating for sensor degradation.
6.3 COMPOUND COMPLEXITY
Real-world odors can consist of dozens or even hundreds of volatile
compounds simultaneously.
It remains unclear how accurately an LLM can interpret increasingly
complex molecular mixtures, particularly when:
• Compounds interact perceptually
• Concentrations vary significantly
• Mixture effects are nonlinear
• The combination is poorly represented in training data
6.4 CONCENTRATION AND PERCEPTION
The mere presence of a compound does not necessarily determine its
perceptual importance.
Human olfaction is influenced by:
• Concentration
• Odor thresholds
• Molecular interactions
• Mixture effects
• Individual differences in perception
Therefore, a future MOA system would likely need more than a simple
binary list of detected compounds.
6.5 LLM HALLUCINATION RISK
LLMs may produce confident but incorrect interpretations, particularly
when presented with:
• Unusual molecular combinations
• Synthetic compounds
• Incomplete measurements
• Conflicting molecular profiles
An operational MOA system would therefore require mechanisms such as:
• Uncertainty estimation
• Chemical verification
• Retrieval-augmented generation
• Structured validation
• Sensor confidence propagation
6.6 NO END-TO-END EMPIRICAL VALIDATION
The proof of concept described in Section 4 evaluates only the
semantic reasoning component using manually constructed input.
No complete end-to-end experiment involving real-time molecular
sensing, automated compound identification, structured input
generation, and blind scent identification has been conducted.
Consequently, the feasibility of the complete MOA architecture remains
an open empirical question.
7. FUTURE WORK
The primary direction for future work is empirical validation.
A minimal viable MOA pipeline could be constructed using:
• Commercially available VOC sensor arrays
• Portable spectrometry hardware
• Open-source molecular sensing platforms
A prototype system could follow this architecture:
┌─────────────────────────┐
│ Physical Environment │
└────────────┬────────────┘
│
▼
┌─────────────────────────┐
│ Molecular Sensor │
└────────────┬────────────┘
│
▼
┌─────────────────────────┐
│ VOC Detection │
└────────────┬────────────┘
│
▼
┌─────────────────────────┐
│ Compound Identification│
└────────────┬────────────┘
│
▼
┌─────────────────────────┐
│ Structured Molecular │
│ Representation │
└────────────┬────────────┘
│
▼
┌─────────────────────────┐
│ LLM │
└────────────┬────────────┘
│
▼
┌─────────────────────────┐
│ Olfactory Interpretation│
└─────────────────────────┘
7.1 EXPERIMENTAL EVALUATION
Controlled experiments could be conducted using a defined scent corpus
containing known molecular compositions.
Potential evaluation metrics could include:
• Scent identification accuracy
• Compound-to-scent reasoning accuracy
• Robustness to concentration changes
• Performance on unseen scent combinations
• False-positive rate
• Confidence calibration
• Agreement with human olfactory assessments
7.2 RETRIEVAL-AUGMENTED GENERATION
A second research direction would investigate whether
Retrieval-Augmented Generation (RAG) over chemical databases can
improve molecular interpretation.
A retrieval system could provide the LLM with verified information
about:
• Molecular structures
• Odor descriptors
• Concentration thresholds
• Known applications
• Chemical properties
• Known associations between compounds and odors
This could reduce hallucination and improve factual grounding.
7.3 SPECIALIZED FINE-TUNING
Another direction would be evaluating whether fine-tuning on
specialized:
• Olfactory chemistry literature
• Experimental odor datasets
• Molecular-to-odor mappings
• Human olfactory assessments
provides meaningful improvements over general-purpose LLMs.
7.4 HYBRID MOLECULAR ENCODERS
Future research could investigate whether MOA requires an LLM at every
stage of interpretation.
A possible alternative would be a hybrid architecture:
Molecular Sensor
│
▼
Molecular Encoder
│
▼
Chemical Representation
│
▼
LLM
│
▼
Semantic Reasoning
│
▼
Natural-Language Output
8. CONCLUSION
This paper presented the Molecular Olfaction Architecture (MOA), a
conceptual framework for extending LLM-based perception into the
olfactory domain.
The central proposal is that an AI system may not necessarily need to
learn smell entirely from scratch.
Instead:
A molecular sensing layer could convert real-world odors into
structured chemical information, while an LLM could use its
existing chemical and semantic knowledge to interpret that
information.
The informal proof of concept presented in this paper suggests that
the reasoning layer is technically plausible: an existing
general-purpose LLM can receive a list of chemical compounds and
generate a coherent interpretation of the associated olfactory
profile.
However, this demonstration does not validate the complete
architecture.
The major unresolved challenge is the sensory interface itself:
Developing affordable, portable, reliable, and sufficiently
precise molecular detection hardware capable of functioning
as an olfactory encoder.
In vision-language systems, cameras provide the sensory bridge between
the physical world and the model.
In audio-language systems, microphones serve a similar function.
MOA proposes that molecular sensors could eventually play an analogous
role for olfaction.
If successful, this could transform smell from a purely descriptive
concept that AI can talk about into a physical sensory modality that AI
can actually measure, interpret, and reason about.
────────────────────────────────────────────────────────────────────────────────
## AUTHOR STATEMENT
This document is a speculative concept paper and has not undergone peer review.
No empirical end-to-end experiments were conducted.
The author declares no conflicts of interest.
No funding was received.
────────────────────────────────────────────────────────────────────────────────