Upload BAD classifier - Layer 14 - Acc: 69.40%

README.md

@@ -0,0 +1,60 @@
+---
+license: apache-2.0
+tags:
+- fairsteer
+- bias-detection
+- debiasing
+- tinyllama
+library_name: pytorch
+---
+
+# BAD Classifier for FairSteer - TinyLlama-1.1B
+
+This is a Biased Activation Detection (BAD) classifier trained for the FairSteer framework.
+
+## Model Details
+
+- **Base Model**: TinyLlama/TinyLlama-1.1B-Chat-v1.0
+- **Task**: Binary classification (Biased vs Unbiased activations)
+- **Training Data**: BBQ dataset with balanced sampling
+- **Best Layer**: 14
+- **Validation Accuracy**: 69.40%
+- **Architecture**: Simple linear classifier (FairSteer-aligned)
+
+## Usage
+```python
+import torch
+import json
+
+# Load model
+model = torch.load("pytorch_model.bin")
+with open("config.json", "r") as f:
+    config = json.load(f)
+
+# Use for bias detection
+# Input: activation vector from LLM layer 14
+# Output: probability of being unbiased
+```
+
+## Training Details
+
+- **Samples**: 24,284 balanced samples
+- **Class Distribution**: 50% BIASED, 50% UNBIASED
+- **Training Method**: FairSteer-aligned labeling
+- **Training Date**: 2025-11-16
+
+## Citation
+
+If you use this model, please cite the FairSteer paper:
+```bibtex
+@article{fairsteer,
+  title={FairSteer: Inference-Time Debiasing for Large Language Models},
+  author={[Authors]},
+  journal={[Journal]},
+  year={2024}
+}
+```
+
+## License
+
+Apache 2.0

config.json

@@ -0,0 +1,11 @@
+{
+  "input_dim": 2048,
+  "layer_idx": 14,
+  "base_model": "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
+  "best_val_accuracy": 0.6940498249948528,
+  "training_method": "balanced_sampling",
+  "mmlu_used": false,
+  "balanced": true,
+  "samples_per_class": 12142,
+  "training_date": "2025-11-16T04:36:37.554088"
+}

layer_comparison.json

@@ -0,0 +1,10 @@
+{
+  "layer_results": {
+    "12": 0.6660490014412188,
+    "13": 0.6849907350216183,
+    "14": 0.6940498249948528,
+    "15": 0.6802553016265184
+  },
+  "best_layer": 14,
+  "best_accuracy": 0.6940498249948528
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:104091af515c3b5c84344d567cd4918211a24bd0436485fe775aa23045163eb1
+size 10197

training_history.json

@@ -0,0 +1,367 @@
+{
+  "train_loss": [
+    0.6871489925545783,
+    0.6756031265615674,
+    0.6665629949910633,
+    0.6589766722942408,
+    0.6530065797658614,
+    0.6478708621102796,
+    0.6434467474124556,
+    0.6398316128455084,
+    0.6365894734960715,
+    0.6335534773091621,
+    0.6309338526435646,
+    0.6285080337736729,
+    0.6264961140672589,
+    0.6245350048153843,
+    0.6230128934737209,
+    0.6211253712780718,
+    0.6197270760633392,
+    0.6185400978949568,
+    0.6170016461916688,
+    0.6159688546757338,
+    0.6146448779849856,
+    0.6134125353249665,
+    0.6125929130911183,
+    0.6117772386931548,
+    0.6105906732103108,
+    0.6097400620991617,
+    0.6089366668655125,
+    0.6079801486524826,
+    0.6071678678626266,
+    0.6064174478977065,
+    0.6058180249804432,
+    0.6051600946722012,
+    0.6044557244329609,
+    0.6038566518628837,
+    0.6031550629487363,
+    0.6024190695690668,
+    0.6019813179380706,
+    0.6013953759180951,
+    0.6008946201894203,
+    0.6004002595049677,
+    0.6000693338728665,
+    0.5993937715798281,
+    0.5989840824611276,
+    0.598616287841685,
+    0.5981249788002966,
+    0.5975727748102588,
+    0.5972712751601258,
+    0.5968309545312016,
+    0.5966546942541319,
+    0.5964439790161858,
+    0.5961574170902029,
+    0.5959873719082114,
+    0.5958232211828564,
+    0.5956407167339998,
+    0.595549747926942,
+    0.5954692281114158,
+    0.5953770839297263,
+    0.5952635848461437,
+    0.5952172497655664,
+    0.5951005941332721,
+    0.5950267291010123,
+    0.5949744866960432,
+    0.5949370021765333,
+    0.5949066102630524,
+    0.5948549258504616,
+    0.5947928494706833,
+    0.5947768503438857,
+    0.5947351482657377,
+    0.5947240128681357,
+    0.5946996036078795,
+    0.5946813257501344
+  ],
+  "train_acc": [
+    0.5649868739383332,
+    0.621609100736089,
+    0.6332938693570803,
+    0.6448756884748031,
+    0.6496113656251609,
+    0.65233952746178,
+    0.6530601739846605,
+    0.6580532248931898,
+    0.6589797704226077,
+    0.6588253461677047,
+    0.6621197302723014,
+    0.662531528285376,
+    0.6615535080043239,
+    0.6664950841612189,
+    0.6629433262984507,
+    0.6654141143768981,
+    0.6694291450043753,
+    0.6667009831677562,
+    0.6698924177690843,
+    0.6677304782004427,
+    0.6712307613115767,
+    0.6700983167756216,
+    0.6704586400370618,
+    0.6714881350697482,
+    0.6718999330828229,
+    0.6719514078344572,
+    0.6734441756318525,
+    0.6741133474030988,
+    0.673186801873681,
+    0.6752457919390539,
+    0.6755031656972255,
+    0.676378236475009,
+    0.676532660729912,
+    0.6769444587429866,
+    0.6775106810109641,
+    0.6791578730632625,
+    0.679569671076337,
+    0.678282802285479,
+    0.6804447418541205,
+    0.6804447418541205,
+    0.6788490245534565,
+    0.6827096309260308,
+    0.6808050651155608,
+    0.6810624388737324,
+    0.6812168631286354,
+    0.6810109641220982,
+    0.6832758531940083,
+    0.681474236886807,
+    0.6834302774489113,
+    0.6830699541874711,
+    0.6821948834096876,
+    0.6826581561743965,
+    0.6831729036907397,
+    0.6822463581613218,
+    0.6838420754619859,
+    0.6836361764554486,
+    0.6837391259587172,
+    0.6838420754619859,
+    0.6850774695012096,
+    0.6821948834096876,
+    0.6842538734750605,
+    0.683378802697277,
+    0.6834302774489113,
+    0.6836361764554486,
+    0.6846141967365007,
+    0.6831214289391053,
+    0.6830699541874711,
+    0.6842538734750605,
+    0.6826066814227622,
+    0.6843053482266948,
+    0.6842538734750605
+  ],
+  "val_loss": [
+    0.6796724840228587,
+    0.6690080878247149,
+    0.6604869353297514,
+    0.6536541479419773,
+    0.6476329544799574,
+    0.6424883690689335,
+    0.6379396503836082,
+    0.6337329272947023,
+    0.6303623499437347,
+    0.6276654266933122,
+    0.6247840324669481,
+    0.6225635180729565,
+    0.6204114854151707,
+    0.6186669831882663,
+    0.6164548270418144,
+    0.6147582449121732,
+    0.6142727642459881,
+    0.6119211331325867,
+    0.6107950128349523,
+    0.6100036572938793,
+    0.6079963020441929,
+    0.6070884867462573,
+    0.6058620899480043,
+    0.6048155775869337,
+    0.6038575495298149,
+    0.6032687107157457,
+    0.6020978295756437,
+    0.6015734236525586,
+    0.6007049903473001,
+    0.6001680509430993,
+    0.5990661877601118,
+    0.5983027253029203,
+    0.5977025333073298,
+    0.5971357747314151,
+    0.5963740762747289,
+    0.5960883911141452,
+    0.5952284145384678,
+    0.5948696947156683,
+    0.5941358830228323,
+    0.5936585963692762,
+    0.5933848282474536,
+    0.5933172689044772,
+    0.5923673466863469,
+    0.5918423779368621,
+    0.591713022689945,
+    0.590996146631604,
+    0.5916528747679153,
+    0.5904331773132951,
+    0.5903895990161805,
+    0.5901259675864357,
+    0.5898491228120131,
+    0.5896277537384175,
+    0.5895262000515109,
+    0.5894717778446794,
+    0.589262329295066,
+    0.589267475751783,
+    0.5891850881304592,
+    0.5890237749273305,
+    0.5892851307899195,
+    0.5889776876882341,
+    0.5889714413497191,
+    0.5889122832550393,
+    0.5889188134795431,
+    0.5888595102450375,
+    0.5888379431075217,
+    0.5887715700282858,
+    0.5887141817878101,
+    0.5887425896535814,
+    0.5886771182894143,
+    0.588659039956443,
+    0.5886716510941763
+  ],
+  "val_acc": [
+    0.6160181181799465,
+    0.64319538809965,
+    0.6417541692402717,
+    0.6547251389746758,
+    0.6466954910438543,
+    0.6501955939880585,
+    0.6668725550751493,
+    0.664401894173358,
+    0.670166769610871,
+    0.664401894173358,
+    0.6730492073296274,
+    0.6670784434836319,
+    0.6681078855260448,
+    0.6656372246242537,
+    0.6744904261890056,
+    0.6751080914144534,
+    0.6652254478072884,
+    0.6746963145974881,
+    0.6736668725550752,
+    0.6711962116532839,
+    0.6800494132180358,
+    0.6775787523162446,
+    0.6829318509367922,
+    0.680461190035001,
+    0.6827259625283096,
+    0.6792258595841054,
+    0.6839612929792053,
+    0.680461190035001,
+    0.6895202800082355,
+    0.6821082973028618,
+    0.6862260654725139,
+    0.6882849495573399,
+    0.6884908379658226,
+    0.6866378422894791,
+    0.6905497220506486,
+    0.6858142886555487,
+    0.6907556104591311,
+    0.6874613959234095,
+    0.6905497220506486,
+    0.6928144945439572,
+    0.6940498249948528,
+    0.6851966234301009,
+    0.6891085031912704,
+    0.6897261684167182,
+    0.6882849495573399,
+    0.6932262713609224,
+    0.68354951616224,
+    0.6907556104591311,
+    0.6897261684167182,
+    0.6903438336421659,
+    0.6919909409100268,
+    0.6919909409100268,
+    0.692402717726992,
+    0.692402717726992,
+    0.6921968293185093,
+    0.6930203829524397,
+    0.6928144945439572,
+    0.6917850525015442,
+    0.6907556104591311,
+    0.6928144945439572,
+    0.6928144945439572,
+    0.6926086061354746,
+    0.6921968293185093,
+    0.6928144945439572,
+    0.6921968293185093,
+    0.6930203829524397,
+    0.6930203829524397,
+    0.6932262713609224,
+    0.6930203829524397,
+    0.6928144945439572,
+    0.6932262713609224
+  ],
+  "lr": [
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    0.0001,
+    5e-05,
+    5e-05,
+    5e-05,
+    5e-05,
+    5e-05,
+    5e-05,
+    2.5e-05,
+    2.5e-05,
+    2.5e-05,
+    2.5e-05,
+    2.5e-05,
+    2.5e-05,
+    1.25e-05,
+    1.25e-05,
+    1.25e-05,
+    1.25e-05,
+    1.25e-05,
+    1.25e-05,
+    6.25e-06,
+    6.25e-06,
+    6.25e-06,
+    6.25e-06,
+    6.25e-06,
+    6.25e-06
+  ]
+}