mirror of
https://github.com/zooai/proofs.git
synced 2026-10-30 14:49:47 +00:00
zeekayandHanzo Dev
638fba4c90
Bylines, paper authorship, correspondence addresses and the founder blurb under each post named an individual. They name the organisation that published the work instead: the brand on posts and pages, the research byline on papers, and research@ in place of a personal address. Both are what these repositories already used elsewhere — nothing new was coined. Co-authors are untouched. Where a paper named someone else alongside, they stay named; only the one byline changed. Bibliography entries, running heads, PDF metadata and affiliations move with it, so a citation still resolves to whoever the cited paper names. Sample and fixture data uses the placeholder name these tests already use elsewhere, so no assertion changed meaning. Co-authored-by: Hanzo Dev <[email protected]>
97 lines
3.5 KiB
Lean4
97 lines
3.5 KiB
Lean4
/-
|
||
Authors: Antje Worring, Zoo Labs Foundation Research
|
||
|
||
Per-User AI Personalization (HIP-0006/HIP-0022)
|
||
|
||
Own Your AI: personalized model adaptation via experience curation.
|
||
Users own their adaptation data, stored at uor.foundation.
|
||
No fine-tuning needed (HLLM conservation: Ψ·Θ = κ).
|
||
|
||
Maps to:
|
||
- HIP-0006: Per-User Fine-Tuning Architecture
|
||
- HIP-0022: Personalized AI — Own Your AI
|
||
- zoo/papers/hllm-training-free-grpo.tex: HLLM foundation
|
||
- zoo/formal/lean/AI/DSO.lean: shared prior aggregation
|
||
|
||
Properties:
|
||
- Ownership: user controls their experience data
|
||
- Privacy: personalization doesn't require sharing raw data
|
||
- Portability: experiences are content-addressed (uor.foundation)
|
||
- Training-free: no gradient computation for personalization
|
||
-/
|
||
|
||
import Mathlib.Data.Nat.Defs
|
||
import Mathlib.Data.List.Basic
|
||
import Mathlib.Tactic
|
||
|
||
namespace AI.Personalization
|
||
|
||
/-- A user's personalization profile -/
|
||
structure UserProfile where
|
||
userId : Nat
|
||
experiences : List Nat -- content hashes of experiences
|
||
domainWeights : List (String × Nat) -- domain preferences
|
||
privacyLevel : Nat -- 0=public, 1=friends, 2=private
|
||
|
||
/-- Personalization state -/
|
||
structure PersonalizationState where
|
||
profile : UserProfile
|
||
contextQuality : Nat -- Ψ in HLLM (Ψ·Θ = κ)
|
||
modelFrozen : Bool -- Θ is never modified
|
||
|
||
/-- Add experience to profile -/
|
||
def addExperience (s : PersonalizationState) (expHash : Nat) : PersonalizationState :=
|
||
{ s with
|
||
profile := { s.profile with experiences := expHash :: s.profile.experiences }
|
||
contextQuality := s.contextQuality + 1 }
|
||
|
||
/-- OWNERSHIP: User controls their data -/
|
||
def isOwnedBy (s : PersonalizationState) (userId : Nat) : Bool :=
|
||
s.profile.userId == userId
|
||
|
||
/-- TRAINING-FREE: Model parameters never change -/
|
||
theorem model_stays_frozen (s : PersonalizationState) (expHash : Nat)
|
||
(h : s.modelFrozen = true) :
|
||
(addExperience s expHash).modelFrozen = true := by
|
||
simp [addExperience, h]
|
||
|
||
/-- CONTEXT IMPROVES: More experiences → better context -/
|
||
theorem context_monotone (s : PersonalizationState) (expHash : Nat) :
|
||
(addExperience s expHash).contextQuality > s.contextQuality := by
|
||
simp only [addExperience]; omega
|
||
|
||
/-- EXPERIENCE COUNT GROWS -/
|
||
theorem experiences_grow (s : PersonalizationState) (expHash : Nat) :
|
||
(addExperience s expHash).profile.experiences.length =
|
||
s.profile.experiences.length + 1 := by
|
||
simp [addExperience, List.length_cons]
|
||
|
||
/-- PRIVACY: Private profiles are not shared -/
|
||
def canShare (s : PersonalizationState) (requesterId : Nat) : Bool :=
|
||
s.profile.privacyLevel == 0 || -- public
|
||
(s.profile.privacyLevel == 1 && requesterId == s.profile.userId) -- simplified
|
||
|
||
theorem private_not_shared (s : PersonalizationState) (requesterId : Nat)
|
||
(h : s.profile.privacyLevel = 2) (_h_diff : requesterId ≠ s.profile.userId) :
|
||
canShare s requesterId = false := by
|
||
simp [canShare, h]
|
||
|
||
/-- PORTABILITY: Experiences are content-addressed hashes.
|
||
Same hash = same experience regardless of platform. -/
|
||
theorem portable_by_hash (exp1 exp2 : Nat) (h : exp1 = exp2) :
|
||
exp1 = exp2 := h
|
||
|
||
/-- ZERO EXPERIENCES: Fresh profile has no personalization -/
|
||
def freshProfile (userId : Nat) : PersonalizationState :=
|
||
{ profile := ⟨userId, [], [], 2⟩ -- private by default
|
||
, contextQuality := 0
|
||
, modelFrozen := true }
|
||
|
||
theorem fresh_zero_context (userId : Nat) :
|
||
(freshProfile userId).contextQuality = 0 := rfl
|
||
|
||
theorem fresh_is_frozen (userId : Nat) :
|
||
(freshProfile userId).modelFrozen = true := rfl
|
||
|
||
end AI.Personalization
|