harshal3099/apex-food-rd-chatml-v3-flavour

Apex Food R&D ChatML v3 — Expanded Ingredients + Flavour & Taste System Design

This v3 dataset extends the Apex Food R&D v2 dataset by adding a dedicated 12th capability:

12. Flavour & Taste System Design

The new capability covers:

• —Indian flavour palette design
• —sweetness modulation
• —bitterness masking systems
• —acid-sweet balance
• —spice-flavour pairing
• —dairy vs water flavour differences
• —natural flavour systems
• —flavour top/middle/base notes
• —flavour release in powders
• —aftertaste control for stevia/monk fruit
• —cocoa/coffee/malt/fruit/spice flavour architectures
• —children vs adults vs diaspora taste preference
• —sensory panel scoring systems
• —hedonic testing
• —JAR scale testing
• —descriptive analysis
• —flavour stability during shelf life
• —flavour oxidation and packaging interaction
• —masking of probiotics, mushrooms, algae, moringa, ashwagandha, pea protein, millets and other functional ingredients

Intended model

Recommended base model: Qwen/Qwen3-4B Model URL: https://huggingface.co/Qwen/Qwen3-4B

Qwen3-4B was selected because its config verifies Qwen3ForCausalLM, it is not a VLM/conditional-generation architecture, it has strong 4B-class quality, and it can later be LoRA fine-tuned on GPU and exported to GGUF.

Dataset size

• —Total examples: 15,000
• —Train: 13,500
• —Validation: 750
• —Test: 750
• —New dedicated flavour/taste examples: 3,000
• —Format: ChatML messages

Capability distribution

• —Ingredient Functionality: 1,982
• —Regulatory Permissibility: 1,983
• —Preservation System Design: 1,357
• —Clean-Label Substitution: 1,357
• —Process Engineering: 939
• —FSSAI & European Standards Health/Nutrition Claims: 939
• —ICMR-NIN Dietary Gap Analysis: 887
• —Organic Certification: 626
• —Fermentation Science: 626
• —Shelf-Life Prediction: 626
• —Texture & Sensory Design: 678
• —Flavour & Taste System Design: 3,000

Ingredient universe

v3 retains the v2 expanded ingredient universe of 137 India-relevant functional/natural/organic ingredients, including millets, pulses, seeds, spices, herbs, Indian fruits, leafy greens, microgreens, mushrooms, algae, fermented ingredients, fibres, sweeteners and probiotic strains.

Regulatory/source grounding

Examples cite and reason from named references including:

• —FSS Act, 2006
• —FSS Food Products Standards and Food Additives Regulations, 2011
• —FSS Labelling and Display Regulations, 2020
• —FSS Advertising and Claims Regulations, 2018
• —FSS Health Supplements, Nutraceuticals, FSDU, FSMP, Prebiotic and Probiotic Food Regulations, 2022
• —Codex GSFA CXS 192-1995
• —Codex CXS 1-1985, CXG 2-1985, CXG 23-1997
• —ICMR-NIN RDA/EAR 2020 and Dietary Guidelines
• —NPOP / APEDA organic standards
• —FAO/WHO probiotic evaluation guidance, 2002
• —ISO 13299 sensory profiling guidance
• —ISO 11136 hedonic consumer testing guidance
• —EU Regulation 1333/2008, 1334/2008, 1924/2006, 2015/2283 as stricter comparison references

Important limitation

This is a synthetic dataset from an expanded curated knowledge base. It is suitable for SFT response style, food formulation reasoning, flavour/sensory design reasoning and regulatory citation behaviour. It is not a substitute for direct legal/regulatory verification against current FSSAI notifications and official PDFs before product launch.