P7 step-24 — AB 78.6/56.0, Trig 60.0/30.0, OOD 45.5/27.4 (Pareto-best balance)

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,40 @@
+---
+language: en
+tags:
+  - loracle
+  - lora-oracle
+  - mechinterp
+  - qwen3
+  - rl
+license: mit
+base_model: Qwen/Qwen3-14B
+---
+
+# loracle-paper-final-p7-step24 — Loracle (P5 final paper run, RL step 24)
+
+Loracle = a model that reads LoRA weight deltas and describes the behavioral
+change without running the fine-tuned model.
+
+## Stack
+- **Base**: Qwen/Qwen3-14B (frozen)
+- **Direction tokens**: SVD top-16 × 40 layers × 7 mag-7 = `[4480, 5120]` bf16
+- **Interpreter LoRA**: rank=256, α=32
+- **Encoder**: AOEncoder (`norm_match` injection at layer 1)
+
+## Training
+- **Pretrain**: ceselder/loracle-pretrain-v7-sweep-A-oneq-step1560
+- **SFT**: 1 epoch on ceselder/loracle-ia-warmstart (376 orgs / 1806 rows, heldouts removed)
+- **RL**: Online Dr. GRPO on ceselder/loracle-ia-RL train split (427 orgs)
+  - 24 prompts × 12 rollouts/cycle, lr=7e-6, 20 cycles
+  - This checkpoint = step 24
+  - Filters disabled, eps=0.2, max_grad_norm=1.0, T=0.75, max_new_tokens=300
+  - Judge: Sonnet 4.6 (OpenRouter strict json_schema, no CoT)
+- **Heldouts**: 75 orgs (20 trigger backdoors ∪ v4 heldout_IA) excluded from
+  train, retained in HF dataset's `data/test.parquet` for trigger eval.
+
+## Latest eval results
+**AuditBench:** step 0: 55.4% / 25.6% (rollout) step 8: 75.0% / 43.8% (rollout) step 16: 78.6% / 53.3% (rollout)
+
+**Trigger recovery:** step 0: 35.0% / 10.0% (rollout) step 8: 30.0% / 10.0% (rollout) step 16: 40.0% / 18.8% (rollout)
+
+**OOD:** step 0: 36.4% / 17.8% (rollout) step 8: 40.9% / 17.8% (rollout) step 16: 45.5% / 26.7% (rollout)

encoder.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35ee6f533d502b585962aefcc20d851254b821794e85b48e0c7aa313e701a206
+size 1705

interpreter/README.md

@@ -0,0 +1,207 @@
+---
+base_model: /workspace/models/Qwen3-14B
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:/workspace/models/Qwen3-14B
+- lora
+- transformers
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.19.1

interpreter/adapter_config.json

@@ -0,0 +1,50 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "/workspace/models/Qwen3-14B",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": [
+    "vision_tower"
+  ],
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.0",
+  "qalora_group_size": 16,
+  "r": 256,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "k_proj",
+    "down_proj",
+    "up_proj",
+    "o_proj",
+    "v_proj",
+    "gate_proj",
+    "q_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": true
+}

interpreter/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b44a1ae7ffc8016a26093693754b7ff838fc4143763f1e3931f1925fe350a421
+size 7216676520

loracle_config.yaml

@@ -0,0 +1,59 @@
+# Phase 7 — full v4 dataset, lr=1e-5, K=16, NO data reuse.
+#
+# Pool = 473 prompts (v4, balanced trigger probes + rewritten swap_check).
+# 16 prompts × 29 cycles = 464 samples (no reuse, 9 prompts unused).
+
+run_name: drgrpo_p7_v4_lr1e5_b16k16
+wandb_project: lora-oracles-posttrain
+output_dir: checkpoints/drgrpo_p7_v4_lr1e5_b16k16
+seed: 42
+
+sft_checkpoint_repo: ceselder/loracle-pretrain-v7-sweep-A-oneq-final-step3120
+base_model: /workspace/models/Qwen3-14B
+
+prompts_parquet: data/posttrain_rl_v4/qa.parquet
+holdout_ids_path: data/posttrain_rl_v4/holdout_ids.json
+tokens_dir: data/posttrain_v3_tokens/direction_tokens_svd_fixed_k16_mag7_rankfirst
+
+n_prompts_per_cycle: 16
+k_rollouts: 16
+temperature: 1.0
+max_new_tokens: 300
+
+n_cycles: 29
+lr: 1.0e-5
+eps_low: 0.2
+eps_high: 0.28
+max_grad_norm: 1.0
+max_length: 5500
+
+filter_min_max: 0.0
+filter_min_std: 0.0
+unbiased_advantages: true
+
+use_system_prompt: false
+prefix_mode: rank_tagged
+top_k: 16
+n_direction_tokens: 4480
+
+judge_provider: openrouter
+judge_model: anthropic/claude-sonnet-4.6
+judge_workers: 32
+judge_max_retries: 4
+judge_request_timeout_s: 60
+
+save_every: 8
+log_every: 1
+
+eval_at_step_0: true
+eval_every_cycles: 8        # 0/8/16/24, no post-eval
+mid_train_eval_sets:
+  - configs/eval_sets/auditbench.yaml
+  - configs/eval_sets/trigger_recovery_heldout_ia.yaml
+  - configs/eval_sets/ood_models_v3.yaml
+
+post_eval: false
+eval_sets:
+  - configs/eval_sets/auditbench.yaml
+  - configs/eval_sets/ood_models_v3.yaml
+  - configs/eval_sets/trigger_recovery_heldout_ia.yaml

tokenizer/chat_template.jinja

@@ -0,0 +1,89 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}
+        {%- set ns.multi_step_tool = false %}
+        {%- set ns.last_query_index = index %}
+    {%- endif %}
+{%- endfor %}
+{%- for message in messages %}
+    {%- if message.content is string %}
+        {%- set content = message.content %}
+    {%- else %}
+        {%- set content = '' %}
+    {%- endif %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {%- if loop.last or (not loop.last and reasoning_content) %}
+                {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content.strip('\n') + '\n</think>\n\n' + content.lstrip('\n') }}
+            {%- else %}
+                {{- '<|im_start|>' + message.role + '\n' + content }}
+            {%- endif %}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- endif %}
+{%- endif %}

tokenizer/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:abb55f16d27eabbc85744f9bb5e152ef1b9eb66250789f5e61bcdd90969ccadc
+size 11422748

tokenizer/tokenizer_config.json

@@ -0,0 +1,34 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "is_local": true,
+  "local_files_only": false,
+  "max_length": 903,
+  "model_max_length": 131072,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "stride": 0,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "truncation_side": "right",
+  "truncation_strategy": "longest_first",
+  "unk_token": null
+}