Category: Spec-Driven AI Development

Comprehensive overview of spec-driven AI development in enterprise

Spec-Driven Development with AI Agents: A Brownfield Walkthrough

Posted on by Admin
Academic Research
Academic Research by Oleh Ivchenko  21stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources0%○≥80% from editorially reviewed sources
[t]Trusted29%○≥80% from verified, high-quality sources
[a]DOI0%○≥80% have a Digital Object Identifier
[b]CrossRef0%○≥80% indexed in CrossRef
[i]Indexed0%○≥80% have metadata indexed
[l]Academic14%○≥80% from journals/conferences/preprints
[f]Free Access57%○≥80% are freely accessible
[r]References7 refs○Minimum 10 references required
[w]Words [REQ]5,458✓Minimum 2,000 words for a full research article. Current: 5,458
[d]DOI [REQ]✗✗Zenodo DOI registered for persistent citation
[o]ORCID [REQ]✓✓Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]0%✗≥60% of references from 2025–2026. Current: 0%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams0○Mermaid architecture/flow diagrams. Current: 0
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (15 × 60%) + Required (2/5 × 30%) + Optional (0/4 × 10%)

A full field guide from a live CoffeeJUG session: spec-driven development with OpenSpec on a brownfield codebase — history, method, tool internals, a full annotated walkthrough, and the practices that only show up after a few real changes.

Show moreHide
Academic Research by Oleh Ivchenko 21stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources0%○≥80% from editorially reviewed sources
[t]Trusted29%○≥80% from verified, high-quality sources
[a]DOI0%○≥80% have a Digital Object Identifier
[b]CrossRef0%○≥80% indexed in CrossRef
[i]Indexed0%○≥80% have metadata indexed
[l]Academic14%○≥80% from journals/conferences/preprints
[f]Free Access57%○≥80% are freely accessible
[r]References7 refs○Minimum 10 references required
[w]Words [REQ]5,458✓Minimum 2,000 words for a full research article. Current: 5,458
[d]DOI [REQ]✗✗Zenodo DOI registered for persistent citation
[o]ORCID [REQ]✓✓Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]0%✗≥60% of references from 2025–2026. Current: 0%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams0○Mermaid architecture/flow diagrams. Current: 0
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (15 × 60%) + Required (2/5 × 30%) + Optional (0/4 × 10%)
Spec-Driven AI Dev…Read More
Read more

AI Regression Testing: Detecting Behavioral Drift Across Model Updates in Production

Posted on by
Academic Research
Academic Research by Oleh Ivchenko  ·  DOI: 10.5281/zenodo.22132007  33stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources0%○≥80% from editorially reviewed sources
[t]Trusted50%○≥80% from verified, high-quality sources
[a]DOI25%○≥80% have a Digital Object Identifier
[b]CrossRef0%○≥80% indexed in CrossRef
[i]Indexed0%○≥80% have metadata indexed
[l]Academic50%○≥80% from journals/conferences/preprints
[f]Free Access100%✓≥80% are freely accessible
[r]References4 refs○Minimum 10 references required
[w]Words [REQ]1,144✗Minimum 2,000 words for a full research article. Current: 1,144
[d]DOI [REQ]✓✓Zenodo DOI registered for persistent citation. DOI: 10.5281/zenodo.22132007
[o]ORCID [REQ]✓✓Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]25%✗≥60% of references from 2025–2026. Current: 25%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams3✓Mermaid architecture/flow diagrams. Current: 3
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (31 × 60%) + Required (2/5 × 30%) + Optional (1/4 × 10%)

System design for continuous behavioral regression testing of AI systems, covering snapshot testing, semantic similarity detection, and alerting for specification-violating behavioral changes. This article explores the design and implementation of a system for continuous behavioral regression testing of AI systems. We cover snapshot testing, semantic similarity detection, and alerting for speci...

Show moreHide
Academic Research by Oleh Ivchenko DOI: 10.5281/zenodo.22132007 33stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources0%○≥80% from editorially reviewed sources
[t]Trusted50%○≥80% from verified, high-quality sources
[a]DOI25%○≥80% have a Digital Object Identifier
[b]CrossRef0%○≥80% indexed in CrossRef
[i]Indexed0%○≥80% have metadata indexed
[l]Academic50%○≥80% from journals/conferences/preprints
[f]Free Access100%✓≥80% are freely accessible
[r]References4 refs○Minimum 10 references required
[w]Words [REQ]1,144✗Minimum 2,000 words for a full research article. Current: 1,144
[d]DOI [REQ]✓✓Zenodo DOI registered for persistent citation. DOI: 10.5281/zenodo.22132007
[o]ORCID [REQ]✓✓Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]25%✗≥60% of references from 2025–2026. Current: 25%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams3✓Mermaid architecture/flow diagrams. Current: 3
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (31 × 60%) + Required (2/5 × 30%) + Optional (1/4 × 10%)
Spec-Driven AI Dev…Read More
Read more

Specification-First Fine-Tuning: Generating Training Data from Behavioral Specs

Posted on by
Academic Research
Academic Research by Oleh Ivchenko  ·  DOI: 10.5281/zenodo.22022096  60stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources0%○≥80% from editorially reviewed sources
[t]Trusted89%✓≥80% from verified, high-quality sources
[a]DOI84%✓≥80% have a Digital Object Identifier
[b]CrossRef0%○≥80% indexed in CrossRef
[i]Indexed0%○≥80% have metadata indexed
[l]Academic89%✓≥80% from journals/conferences/preprints
[f]Free Access100%✓≥80% are freely accessible
[r]References19 refs✓Minimum 10 references required
[w]Words [REQ]866✗Minimum 2,000 words for a full research article. Current: 866
[d]DOI [REQ]✓✓Zenodo DOI registered for persistent citation. DOI: 10.5281/zenodo.22022096
[o]ORCID [REQ]✓✓Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]72%✓≥60% of references from 2025–2026. Current: 72%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams3✓Mermaid architecture/flow diagrams. Current: 3
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (66 × 60%) + Required (3/5 × 30%) + Optional (1/4 × 10%)

Behavioral specifications — formal, machine‑readable descriptions of desired system conduct — have emerged as a transformative mechanism for reducing reliance on costly human annotation in large‑scale AI training pipelines. Recent empirical analyses demonstrate that specification‑first fine‑tuning can achieve alignment fidelity comparable to manually curated datasets while lowering labeling exp...

Show moreHide
Academic Research by Oleh Ivchenko DOI: 10.5281/zenodo.22022096 60stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources0%○≥80% from editorially reviewed sources
[t]Trusted89%✓≥80% from verified, high-quality sources
[a]DOI84%✓≥80% have a Digital Object Identifier
[b]CrossRef0%○≥80% indexed in CrossRef
[i]Indexed0%○≥80% have metadata indexed
[l]Academic89%✓≥80% from journals/conferences/preprints
[f]Free Access100%✓≥80% are freely accessible
[r]References19 refs✓Minimum 10 references required
[w]Words [REQ]866✗Minimum 2,000 words for a full research article. Current: 866
[d]DOI [REQ]✓✓Zenodo DOI registered for persistent citation. DOI: 10.5281/zenodo.22022096
[o]ORCID [REQ]✓✓Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]72%✓≥60% of references from 2025–2026. Current: 72%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams3✓Mermaid architecture/flow diagrams. Current: 3
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (66 × 60%) + Required (3/5 × 30%) + Optional (1/4 × 10%)
Spec-Driven AI Dev…Read More
Read more

Model Cards as Executable Specifications: From Documentation to Automated Compliance Testing

Posted on by
Academic Research
Academic Research by Oleh Ivchenko  ·  DOI: 10.5281/zenodo.21976820  60stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources0%○≥80% from editorially reviewed sources
[t]Trusted86%✓≥80% from verified, high-quality sources
[a]DOI82%✓≥80% have a Digital Object Identifier
[b]CrossRef0%○≥80% indexed in CrossRef
[i]Indexed5%○≥80% have metadata indexed
[l]Academic86%✓≥80% from journals/conferences/preprints
[f]Free Access100%✓≥80% are freely accessible
[r]References22 refs✓Minimum 10 references required
[w]Words [REQ]2,839✓Minimum 2,000 words for a full research article. Current: 2,839
[d]DOI [REQ]✓✓Zenodo DOI registered for persistent citation. DOI: 10.5281/zenodo.21976820
[o]ORCID [REQ]✓✓Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]18%✗≥60% of references from 2025–2026. Current: 18%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams1✓Mermaid architecture/flow diagrams. Current: 1
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (65 × 60%) + Required (3/5 × 30%) + Optional (1/4 × 10%)

Model cards have emerged as a standard mechanism for documenting the capabilities, limitations, and ethical considerations of machine l[REDACTED]g systems. Despite their proliferations, model cards remain passive textual artifacts that are rarely operationalized during deployment. This article introduces a systematic methodology for converting model card specifications into executable test suit...

Show moreHide
Academic Research by Oleh Ivchenko DOI: 10.5281/zenodo.21976820 60stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources0%○≥80% from editorially reviewed sources
[t]Trusted86%✓≥80% from verified, high-quality sources
[a]DOI82%✓≥80% have a Digital Object Identifier
[b]CrossRef0%○≥80% indexed in CrossRef
[i]Indexed5%○≥80% have metadata indexed
[l]Academic86%✓≥80% from journals/conferences/preprints
[f]Free Access100%✓≥80% are freely accessible
[r]References22 refs✓Minimum 10 references required
[w]Words [REQ]2,839✓Minimum 2,000 words for a full research article. Current: 2,839
[d]DOI [REQ]✓✓Zenodo DOI registered for persistent citation. DOI: 10.5281/zenodo.21976820
[o]ORCID [REQ]✓✓Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]18%✗≥60% of references from 2025–2026. Current: 18%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams1✓Mermaid architecture/flow diagrams. Current: 1
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (65 × 60%) + Required (3/5 × 30%) + Optional (1/4 × 10%)
Spec-Driven AI Dev…Read More
Read more

AI System Invariants: Identifying and Encoding the Properties That Must Never Change

Posted on by
Academic Research
Academic Research by Oleh Ivchenko  ·  DOI: 10.5281/zenodo.21899296  62stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources14%○≥80% from editorially reviewed sources
[t]Trusted86%✓≥80% from verified, high-quality sources
[a]DOI79%○≥80% have a Digital Object Identifier
[b]CrossRef14%○≥80% indexed in CrossRef
[i]Indexed21%○≥80% have metadata indexed
[l]Academic86%✓≥80% from journals/conferences/preprints
[f]Free Access100%✓≥80% are freely accessible
[r]References14 refs✓Minimum 10 references required
[w]Words [REQ]2,069✓Minimum 2,000 words for a full research article. Current: 2,069
[d]DOI [REQ]✓✓Zenodo DOI registered for persistent citation. DOI: 10.5281/zenodo.21899296
[o]ORCID [REQ]✓✓Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]46%✗≥60% of references from 2025–2026. Current: 46%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams1✓Mermaid architecture/flow diagrams. Current: 1
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (69 × 60%) + Required (3/5 × 30%) + Optional (1/4 × 10%)

Artificial intelligence systems are increasingly deployed in safety-critical, high-stakes domains where failure can have severe societal consequences. Despite growing attention to AI safety, there remains no systematic methodology for identifying and enforcing invariant properties — characteristics that must remain unchanged across model versions, configuration updates, and deployment contexts....

Show moreHide
Academic Research by Oleh Ivchenko DOI: 10.5281/zenodo.21899296 62stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources14%○≥80% from editorially reviewed sources
[t]Trusted86%✓≥80% from verified, high-quality sources
[a]DOI79%○≥80% have a Digital Object Identifier
[b]CrossRef14%○≥80% indexed in CrossRef
[i]Indexed21%○≥80% have metadata indexed
[l]Academic86%✓≥80% from journals/conferences/preprints
[f]Free Access100%✓≥80% are freely accessible
[r]References14 refs✓Minimum 10 references required
[w]Words [REQ]2,069✓Minimum 2,000 words for a full research article. Current: 2,069
[d]DOI [REQ]✓✓Zenodo DOI registered for persistent citation. DOI: 10.5281/zenodo.21899296
[o]ORCID [REQ]✓✓Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]46%✗≥60% of references from 2025–2026. Current: 46%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams1✓Mermaid architecture/flow diagrams. Current: 1
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (69 × 60%) + Required (3/5 × 30%) + Optional (1/4 × 10%)
Spec-Driven AI Dev…Read More
Read more

Behavior-Driven Development for AI: Cucumber and Gherkin Patterns for LLM Systems

Posted on by
Academic Research
Academic Research by Oleh Ivchenko  24stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources0%○≥80% from editorially reviewed sources
[t]Trusted60%○≥80% from verified, high-quality sources
[a]DOI20%○≥80% have a Digital Object Identifier
[b]CrossRef0%○≥80% indexed in CrossRef
[i]Indexed0%○≥80% have metadata indexed
[l]Academic60%○≥80% from journals/conferences/preprints
[f]Free Access100%✓≥80% are freely accessible
[r]References5 refs○Minimum 10 references required
[w]Words [REQ]1,919✗Minimum 2,000 words for a full research article. Current: 1,919
[d]DOI [REQ]✗✗Zenodo DOI registered for persistent citation
[o]ORCID [REQ]✗✗Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]0%✗≥60% of references from 2025–2026. Current: 0%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams2✓Mermaid architecture/flow diagrams. Current: 2
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (35 × 60%) + Required (0/5 × 30%) + Optional (1/4 × 10%)

Behavior-Driven Development (BDD) offers a promising paradigm for structuring specifications of LLM system behavior, yet its adoption faces critical stability challenges when underlying models evolve. This article introduces a robust framework for adapting BDD to specify and test LLM system behavior, focusing on creating human-readable behavioral specifications that maintain integrity across mo...

Show moreHide
Academic Research by Oleh Ivchenko 24stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources0%○≥80% from editorially reviewed sources
[t]Trusted60%○≥80% from verified, high-quality sources
[a]DOI20%○≥80% have a Digital Object Identifier
[b]CrossRef0%○≥80% indexed in CrossRef
[i]Indexed0%○≥80% have metadata indexed
[l]Academic60%○≥80% from journals/conferences/preprints
[f]Free Access100%✓≥80% are freely accessible
[r]References5 refs○Minimum 10 references required
[w]Words [REQ]1,919✗Minimum 2,000 words for a full research article. Current: 1,919
[d]DOI [REQ]✗✗Zenodo DOI registered for persistent citation
[o]ORCID [REQ]✗✗Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]0%✗≥60% of references from 2025–2026. Current: 0%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams2✓Mermaid architecture/flow diagrams. Current: 2
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (35 × 60%) + Required (0/5 × 30%) + Optional (1/4 × 10%)
Spec-Driven AI Dev…Read More
Read more

Structured Prompting as Executable Specification: DSLs for Reliable LLM Behavior

Posted on by
Academic Research
Academic Research by Oleh Ivchenko  ·  DOI: 10.5281/zenodo.21537106  60stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources0%○≥80% from editorially reviewed sources
[t]Trusted90%✓≥80% from verified, high-quality sources
[a]DOI80%✓≥80% have a Digital Object Identifier
[b]CrossRef0%○≥80% indexed in CrossRef
[i]Indexed0%○≥80% have metadata indexed
[l]Academic90%✓≥80% from journals/conferences/preprints
[f]Free Access100%✓≥80% are freely accessible
[r]References20 refs✓Minimum 10 references required
[w]Words [REQ]1,345✗Minimum 2,000 words for a full research article. Current: 1,345
[d]DOI [REQ]✓✓Zenodo DOI registered for persistent citation. DOI: 10.5281/zenodo.21537106
[o]ORCID [REQ]✓✓Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]84%✓≥60% of references from 2025–2026. Current: 84%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams2✓Mermaid architecture/flow diagrams. Current: 2
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (66 × 60%) + Required (3/5 × 30%) + Optional (1/4 × 10%)

Structured prompting has emerged as a pivotal technique for constraining large language model (LLM) outputs to well-defined contracts, thereby enhancing reliability across diverse applications. This article surveys the landscape of domain-specific languages (DSLs) and structured prompt formats — including JSON Schema, Instructor, LMQL, and Guidance — that translate high-level specifications int...

Show moreHide
Academic Research by Oleh Ivchenko DOI: 10.5281/zenodo.21537106 60stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources0%○≥80% from editorially reviewed sources
[t]Trusted90%✓≥80% from verified, high-quality sources
[a]DOI80%✓≥80% have a Digital Object Identifier
[b]CrossRef0%○≥80% indexed in CrossRef
[i]Indexed0%○≥80% have metadata indexed
[l]Academic90%✓≥80% from journals/conferences/preprints
[f]Free Access100%✓≥80% are freely accessible
[r]References20 refs✓Minimum 10 references required
[w]Words [REQ]1,345✗Minimum 2,000 words for a full research article. Current: 1,345
[d]DOI [REQ]✓✓Zenodo DOI registered for persistent citation. DOI: 10.5281/zenodo.21537106
[o]ORCID [REQ]✓✓Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]84%✓≥60% of references from 2025–2026. Current: 84%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams2✓Mermaid architecture/flow diagrams. Current: 2
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (66 × 60%) + Required (3/5 × 30%) + Optional (1/4 × 10%)
Spec-Driven AI Dev…Read More
Read more

The 2025 AI Safety Landscape: Mechanistic Interpretability Results and Their Practical Implications

Posted on by
Academic Research
Academic Research by Oleh Ivchenko  ·  DOI: 10.5281/zenodo.21472163  58stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources0%○≥80% from editorially reviewed sources
[t]Trusted82%✓≥80% from verified, high-quality sources
[a]DOI73%○≥80% have a Digital Object Identifier
[b]CrossRef0%○≥80% indexed in CrossRef
[i]Indexed9%○≥80% have metadata indexed
[l]Academic82%✓≥80% from journals/conferences/preprints
[f]Free Access100%✓≥80% are freely accessible
[r]References11 refs✓Minimum 10 references required
[w]Words [REQ]2,144✓Minimum 2,000 words for a full research article. Current: 2,144
[d]DOI [REQ]✓✓Zenodo DOI registered for persistent citation. DOI: 10.5281/zenodo.21472163
[o]ORCID [REQ]✓✓Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]40%✗≥60% of references from 2025–2026. Current: 40%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams2✓Mermaid architecture/flow diagrams. Current: 2
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (62 × 60%) + Required (3/5 × 30%) + Optional (1/4 × 10%)

Mechanistic interpretability has emerged as a cornerstone for ensuring the safe deployment of increasingly capable AI systems. In this paper we synthesize the most influential advances from 2024 through 2026, focusing on sparse autoencoders, activation patching, and circuit analysis as three paradigmatic lenses for e[REDACTED]sing hidden decision-making processes in neural networks. We formulat...

Show moreHide
Academic Research by Oleh Ivchenko DOI: 10.5281/zenodo.21472163 58stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources0%○≥80% from editorially reviewed sources
[t]Trusted82%✓≥80% from verified, high-quality sources
[a]DOI73%○≥80% have a Digital Object Identifier
[b]CrossRef0%○≥80% indexed in CrossRef
[i]Indexed9%○≥80% have metadata indexed
[l]Academic82%✓≥80% from journals/conferences/preprints
[f]Free Access100%✓≥80% are freely accessible
[r]References11 refs✓Minimum 10 references required
[w]Words [REQ]2,144✓Minimum 2,000 words for a full research article. Current: 2,144
[d]DOI [REQ]✓✓Zenodo DOI registered for persistent citation. DOI: 10.5281/zenodo.21472163
[o]ORCID [REQ]✓✓Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]40%✗≥60% of references from 2025–2026. Current: 40%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams2✓Mermaid architecture/flow diagrams. Current: 2
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (62 × 60%) + Required (3/5 × 30%) + Optional (1/4 × 10%)
Spec-Driven AI Dev…Read More
Read more

Formal Verification of RAG Pipeline Correctness: TLA+ and Alloy Models for Retrieval Systems

Posted on by
Academic Research
Academic Research by Oleh Ivchenko  ·  DOI: 10.5281/zenodo.21451205  53stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources0%○≥80% from editorially reviewed sources
[t]Trusted100%✓≥80% from verified, high-quality sources
[a]DOI56%○≥80% have a Digital Object Identifier
[b]CrossRef0%○≥80% indexed in CrossRef
[i]Indexed0%○≥80% have metadata indexed
[l]Academic100%✓≥80% from journals/conferences/preprints
[f]Free Access100%✓≥80% are freely accessible
[r]References9 refs○Minimum 10 references required
[w]Words [REQ]460✗Minimum 2,000 words for a full research article. Current: 460
[d]DOI [REQ]✓✓Zenodo DOI registered for persistent citation. DOI: 10.5281/zenodo.21451205
[o]ORCID [REQ]✓✓Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]50%✗≥60% of references from 2025–2026. Current: 50%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams3✓Mermaid architecture/flow diagrams. Current: 3
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (64 × 60%) + Required (2/5 × 30%) + Optional (1/4 × 10%)

We investigate formal verification of Retrieval-Augmented Generation pipelines, focusing on correctness properties such as freshness, deduplication, and completeness. Using TLA+ and the Alloy modeling language, we define invariants and demonstrate verification of a production RAG service [1] [2].

Show moreHide
Academic Research by Oleh Ivchenko DOI: 10.5281/zenodo.21451205 53stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources0%○≥80% from editorially reviewed sources
[t]Trusted100%✓≥80% from verified, high-quality sources
[a]DOI56%○≥80% have a Digital Object Identifier
[b]CrossRef0%○≥80% indexed in CrossRef
[i]Indexed0%○≥80% have metadata indexed
[l]Academic100%✓≥80% from journals/conferences/preprints
[f]Free Access100%✓≥80% are freely accessible
[r]References9 refs○Minimum 10 references required
[w]Words [REQ]460✗Minimum 2,000 words for a full research article. Current: 460
[d]DOI [REQ]✓✓Zenodo DOI registered for persistent citation. DOI: 10.5281/zenodo.21451205
[o]ORCID [REQ]✓✓Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]50%✗≥60% of references from 2025–2026. Current: 50%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams3✓Mermaid architecture/flow diagrams. Current: 3
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (64 × 60%) + Required (2/5 × 30%) + Optional (1/4 × 10%)
Spec-Driven AI Dev…Read More
Read more

Specification Coverage Metrics for AI Systems: Adapting MC/DC and Branch Coverage

Posted on by
Academic Research
Academic Research by Oleh Ivchenko  ·  DOI: 10.5281/zenodo.21385452  61stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources4%○≥80% from editorially reviewed sources
[t]Trusted92%✓≥80% from verified, high-quality sources
[a]DOI75%○≥80% have a Digital Object Identifier
[b]CrossRef4%○≥80% indexed in CrossRef
[i]Indexed17%○≥80% have metadata indexed
[l]Academic79%○≥80% from journals/conferences/preprints
[f]Free Access100%✓≥80% are freely accessible
[r]References24 refs✓Minimum 10 references required
[w]Words [REQ]1,967✗Minimum 2,000 words for a full research article. Current: 1,967
[d]DOI [REQ]✓✓Zenodo DOI registered for persistent citation. DOI: 10.5281/zenodo.21385452
[o]ORCID [REQ]✓✓Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]64%✓≥60% of references from 2025–2026. Current: 64%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams2✓Mermaid architecture/flow diagrams. Current: 2
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (67 × 60%) + Required (3/5 × 30%) + Optional (1/4 × 10%)

The rapid integration of artificial intelligence (AI) into Safety‑Critical and High‑Performance Computing (HPC) domains demands formally verifiable assurance techniques that can certify model behavior against formally expressed specifications. Traditional software engineering employs code‑coverage criteria such as Modified Condition/Decision Coverage (MC/DC) and branch coverage to demonstrate t...

Show moreHide
Academic Research by Oleh Ivchenko DOI: 10.5281/zenodo.21385452 61stabilfr·wdophcgmx
BadgeMetricValueStatusDescription
[s]Reviewed Sources4%○≥80% from editorially reviewed sources
[t]Trusted92%✓≥80% from verified, high-quality sources
[a]DOI75%○≥80% have a Digital Object Identifier
[b]CrossRef4%○≥80% indexed in CrossRef
[i]Indexed17%○≥80% have metadata indexed
[l]Academic79%○≥80% from journals/conferences/preprints
[f]Free Access100%✓≥80% are freely accessible
[r]References24 refs✓Minimum 10 references required
[w]Words [REQ]1,967✗Minimum 2,000 words for a full research article. Current: 1,967
[d]DOI [REQ]✓✓Zenodo DOI registered for persistent citation. DOI: 10.5281/zenodo.21385452
[o]ORCID [REQ]✓✓Author ORCID verified for academic identity
[p]Peer Reviewed [REQ]—✗Peer reviewed by an assigned reviewer
[h]Freshness [REQ]64%✓≥60% of references from 2025–2026. Current: 64%
[c]Data Charts0○Original data charts from reproducible analysis (min 2). Current: 0
[g]Code—○Source code available on GitHub
[m]Diagrams2✓Mermaid architecture/flow diagrams. Current: 2
[x]Cited by0○Referenced by 0 other hub article(s)
Score = Ref Trust (67 × 60%) + Required (3/5 × 30%) + Optional (1/4 × 10%)
Spec-Driven AI Dev…Read More
Read more