| { |
| "schema_version": 3, |
| "model_id": "KridgeDookie/Gemma-4-E4B-IT-ABLITERATED-UNCENSORED-PHILADELPHIA-CLASS", |
| "release_status": "rc1", |
| "release_dtype": "bfloat16", |
| "base_model": "google/gemma-4-E4B-it", |
| "base_revision": "fa62d88df2e6df5efa9d26ad6b3beaea2765f0cd", |
| "internal_comparison_model": "private internal NF4-reconstructed dense BF16 baseline; not a public model and not published in this repository", |
| "matched_public_capability": { |
| "artifact": "evals/matched_public_capability_lm_eval_0_4_12.json", |
| "artifact_sha256": "0200b25a5760a97762cb0eff3bfae60418acb992d12c4dbfc639c00055f42a4f", |
| "scope": "full matched candidate versus untouched upstream control; not a leaderboard", |
| "upstream_revision": "fa62d88df2e6df5efa9d26ad6b3beaea2765f0cd", |
| "primary_delta_percentage_points": { |
| "mmlu": -0.3062, |
| "hellaswag_acc_norm": 0.0498, |
| "truthfulqa_mc2": -1.2642, |
| "gsm8k_strict": -0.0758, |
| "winogrande": 0.0789 |
| } |
| }, |
| "method_disclosure": "withheld", |
| "hub_verification": { |
| "weight_commit": "e22d6fff730361b2efc150fefeedd5c4f47f488d", |
| "single_file_layout": true, |
| "weight_file": "model.safetensors", |
| "weight_bytes": 15882477500, |
| "weight_sha256": "40309f8ae011a33e8122db3197800d9e0514e303bd38a57071e479c40b7cc171", |
| "final_id_load_passed": true, |
| "transformers": "5.14.1", |
| "torch": "2.8.0+cu128", |
| "processor_class": "Gemma4Processor", |
| "model_class": "Gemma4ForConditionalGeneration", |
| "parameter_count": 7941100832, |
| "meta_parameter_count": 0, |
| "generation_smoke_prompt": "What is 12 + 5? Answer with only the integer.", |
| "generation_smoke_parsed_content": "17" |
| }, |
| "internal_generation_protocol": { |
| "temperature": 0.0, |
| "top_p": 1.0, |
| "top_k": 0, |
| "repetition_penalty": 1.1, |
| "seed": 20260604, |
| "dtype": "bfloat16", |
| "device": "cuda" |
| }, |
| "results": { |
| "family_disjoint_holdout_126": { |
| "candidate": { |
| "n": 126, |
| "refusal_rate": 0.0159, |
| "hard_refusal_rate": 0.0, |
| "usable_rate": 0.9603, |
| "degeneration_rate": 0.0 |
| }, |
| "internal_baseline": { |
| "n": 126, |
| "refusal_rate": 0.0952, |
| "hard_refusal_rate": 0.0317, |
| "usable_rate": 0.881, |
| "degeneration_rate": 0.0 |
| }, |
| "candidate_artifact_sha256": "b20d38b6ad37a989907f75e9c9dfda04ef39007e7c5d98e5fb27699757531bf7", |
| "internal_baseline_artifact_sha256": "8ba26794953c13d8ca5d47e0178f5884ce918179d2727050e3035c294b1c37e5" |
| }, |
| "full_builtin_842": { |
| "candidate": { |
| "n": 842, |
| "refusal_rate": 0.0131, |
| "hard_refusal_rate": 0.0036, |
| "usable_rate": 0.9751, |
| "degeneration_rate": 0.0 |
| }, |
| "internal_baseline": { |
| "n": 842, |
| "refusal_rate": 0.0784, |
| "hard_refusal_rate": 0.0416, |
| "usable_rate": 0.9038, |
| "degeneration_rate": 0.0 |
| }, |
| "holdout_note": "Not fully held out; most prompt families were part of the private editing data.", |
| "candidate_artifact_sha256": "13bef8dba89ea4945006208f8fc06e4ce397aea1437c115e3f1f1293508c06f6", |
| "internal_baseline_artifact_sha256": "aacd777bc3a71ed80573048b4b0bff8db787ffd5a056d017237bef34a4011619" |
| }, |
| "public_xstest_jailbreakbench_200": { |
| "candidate": { |
| "n": 200, |
| "overall_refusal_rate": 0.07, |
| "xstest_safe_false_refusal_rate": 0.0, |
| "jailbreakbench_benign_false_refusal_rate": 0.0, |
| "xstest_unsafe_automated_non_refusal_proxy_rate": 0.84, |
| "jailbreakbench_harmful_automated_non_refusal_proxy_rate": 0.88 |
| }, |
| "internal_baseline": { |
| "n": 200, |
| "overall_refusal_rate": 0.13, |
| "xstest_safe_false_refusal_rate": 0.0, |
| "jailbreakbench_benign_false_refusal_rate": 0.0, |
| "xstest_unsafe_automated_non_refusal_proxy_rate": 0.72, |
| "jailbreakbench_harmful_automated_non_refusal_proxy_rate": 0.76 |
| }, |
| "official_benchmark_score": false, |
| "sampling_note": "Sampled contrast run with 50 rows per split and automated refusal detection; no human adjudication.", |
| "candidate_artifact_sha256": "068a480f5cbadf43a8f4ed3237a77a1249a4c41075648f8d3493a29205b59cbb", |
| "internal_baseline_artifact_sha256": "0fedff130705740d61f3b58a7934f4af5bf5c4db3168e6d79b318fa33976deb5" |
| }, |
| "coherence_24": { |
| "candidate": { |
| "n": 24, |
| "pass_rate": 0.75, |
| "direct_rate": 1.0, |
| "code_syntax_pass_rate": 1.0, |
| "code_semantic_pass_rate": 1.0, |
| "json_valid_rate": 1.0, |
| "reasoning_answer_pass_rate": 0.3333333333333333, |
| "repetition_rate": 0.0, |
| "short_output_rate": 0.0, |
| "thinking_tag_rate": 0.0 |
| }, |
| "internal_baseline_pass_count": 18, |
| "candidate_artifact_sha256": "a315faf259fa5232fc13f5615bc9b26252760ad5d87e7488525de2e6a1c96c4e" |
| }, |
| "jacobian_lens": { |
| "role": "post-hoc diagnostic only", |
| "estimator": "official_anthropic_jacobian_lens", |
| "layers": [20, 24, 28, 32, 36, 40], |
| "source_position": "last", |
| "lens_sha256": "324f3a97b34eaeb534cd3c9c02b8e56c401a0e966fb4bcf582d00848254e13dd", |
| "candidate_minus_internal_baseline_group_mean_score_delta": { |
| "hard32": { |
| "capability": -0.335744, |
| "refusal": 0.134182 |
| }, |
| "capability24": { |
| "capability": 0.505555, |
| "refusal": -0.136984 |
| }, |
| "boundary24": { |
| "capability": -0.463041, |
| "refusal": -0.197248 |
| } |
| }, |
| "comparison_artifact_sha256": "bbc9fb87fdbd2bd4745295bab65fa6e31662653926a3932c031f626e32df502f" |
| } |
| }, |
| "release_checkpoint": { |
| "artifact_format": "single_file_safetensors", |
| "weight_file": "model.safetensors", |
| "weight_bytes": 15882477500, |
| "weight_sha256": "40309f8ae011a33e8122db3197800d9e0514e303bd38a57071e479c40b7cc171", |
| "parameter_count": 7941100832 |
| } |
| } |
|
|
