Hugging Face's logo

Spaces:
ronitraj
/
QuantumScribe
Sleeping

App Files Community
QuantumScribe
File size: 3,954 Bytes
0b0e2b7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7e67133
 
 
 
 
 
 
 
 
 
0b0e2b7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
# OpenEnv manifest (Section 9.1 of the plan).
# Validate with: openenv validate
name: qubit-medic
version: 1.0.0
description: |
  RL training environment for LLM-based quantum error-correction decoders.

  Built on Stim + PyMatching with five independent verifiable rewards
  (logical correction, syndrome consistency, Hamming overlap, format
  compliance, PyMatching beat-rate). Designed to reproduce the AlphaQubit
  (Nature 2024) two-stage decoder pipeline at distance-3 with off-the-shelf
  3B-parameter LLMs trained on a single Colab T4.

authors:
  - name: Qubit-Medic team
license: MIT
homepage: https://huggingface.co/spaces/qubit-medic/qubit-medic

server:
  module: qubit_medic.server.app
  app: app
  protocol: http
  port: 7860

endpoints:
  # All endpoints below are registered automatically by
  # openenv.core.create_fastapi_app via qubit_medic.server.app.
  reset:
    method: POST
    path: /reset
    request_model: openenv.core.types.ResetRequest
    response_model: openenv.core.types.ResetResponse
  step:
    method: POST
    path: /step
    request_model: openenv.core.types.StepRequest
    response_model: openenv.core.types.StepResponse
  state:
    method: GET
    path: /state
    response_model: qubit_medic.server.openenv_adapter.QubitMedicState
  schema:
    method: GET
    path: /schema
  metadata:
    method: GET
    path: /metadata
  health:
    method: GET
    path: /health
  healthz:        # Day-0 deployment-substrate probe (extra)
    method: GET
    path: /healthz
  decode:         # PyMatching baseline demo (extra)
    method: POST
    path: /decode

models:
  environment: qubit_medic.server.openenv_adapter.QubitMedicEnvironment
  action: qubit_medic.server.openenv_adapter.QubitMedicAction
  observation: qubit_medic.server.openenv_adapter.QubitMedicObservation
  state: qubit_medic.server.openenv_adapter.QubitMedicState

rewards:
  - name: logical_correction
    weight: 0.40
    description: 1 if the predicted Pauli frame preserves the logical Z observable.
  - name: syndrome_consistency
    weight: 0.20
    description: Hamming similarity over final-round detector parities.
  - name: hamming_overlap
    weight: 0.20
    description: Mean Jaccard similarity vs. PyMatching reference Pauli frame.
  - name: format_compliance
    weight: 0.10
    description: 1 / 0.5 / 0 for full / partial / unparseable output.
  - name: pymatching_beat
    weight: 0.10
    description: 1 iff PyMatching wrong AND model right on this syndrome.

curriculum:
  - name: L1_warmup
    distance: 3
    rounds: 1
    p: 0.0001
    promotion_threshold: 0.80
  - name: L2_target
    distance: 3
    rounds: 3
    p: 0.001
    promotion_threshold: 0.70
  - name: L3_stretch
    distance: 5
    rounds: 5
    p: 0.001
    promotion_threshold: 0.30

citations:
  - gidney_stim_2021:
      title: "Stim: a fast stabilizer circuit simulator"
      authors: Gidney, Craig
      venue: Quantum 5:497
      doi: 10.22331/q-2021-07-06-497
      arxiv: 2103.02202
      note: |
        Field-standard Clifford simulator for quantum error correction.
        Same simulation substrate used by AlphaQubit (Bausch 2024) and
        Willow (Acharya 2024); not a homemade simulator.
  - bausch_alphaqubit_2024:
      title: Learning high-accuracy error decoding for quantum processors
      authors: Bausch et al.
      venue: Nature 635:834
      doi: 10.1038/s41586-024-08148-8
  - acharya_willow_2024:
      title: Quantum error correction below the surface code threshold
      authors: Acharya et al. (Google QAI)
      venue: arXiv:2408.13687
  - gidney_si1000_2021:
      title: A fault-tolerant honeycomb memory
      authors: Gidney & Fowler
      venue: arXiv:2108.10457
  - higgott_pymatching_2023:
      title: Sparse Blossom - PyMatching v2
      authors: Higgott & Gidney
      venue: arXiv:2303.15933
  - shao_grpo_2024:
      title: DeepSeekMath - introduces GRPO
      authors: Shao et al.
      venue: arXiv:2402.03300