Upload folder using huggingface_hub

LICENSE

@@ -0,0 +1,192 @@
+                                 Apache License Version 2.0, January 2004
+                                 http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction, and
+      distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by the
+      copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all other
+      entities that control, are controlled by, or are under common control with
+      that entity. For the purposes of this definition, "control" means (i) the
+      power, direct or indirect, to cause the direction or management of such
+      entity, whether by contract or otherwise, or (ii) ownership of fifty
+      percent (50%) or more of the outstanding shares, or (iii) beneficial
+      ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity exercising
+      permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation source,
+      and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical transformation
+      or translation of a Source form, including but not limited to compiled
+      object code, generated documentation, and conversions to other media
+      types.
+
+      "Work" shall mean the work of authorship, whether in Source or Object
+      form, made available under the License, as indicated by a copyright notice
+      that is included in or attached to the work (an example is provided in the
+      Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object form,
+      that is based on (or derived from) the Work and for which the editorial
+      revisions, annotations, elaborations, or other modifications represent, as
+      a whole, an original work of authorship. For the purposes of this License,
+      Derivative Works shall not include works that remain separable from, or
+      merely link (or bind by name) to the interfaces of, the Work and
+      Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including the original
+      version of the Work and any modifications or additions to that Work or
+      Derivative Works thereof, that is intentionally submitted to Licensor for
+      inclusion in the Work by the copyright owner or by an individual or Legal
+      Entity authorized to submit on behalf of the copyright owner. For the
+      purposes of this definition, "submitted" means any form of electronic,
+      verbal, or written communication sent to the Licensor or its
+      representatives, including but not limited to communication on electronic
+      mailing lists, source code control systems, and issue tracking systems
+      that are managed by, or on behalf of, the Licensor for the purpose of
+      discussing and improving the Work, but excluding communication that is
+      conspicuously marked or otherwise designated in writing by the copyright
+      owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity on
+      behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of this
+      License, each Contributor hereby grants to You a perpetual, worldwide,
+      non-exclusive, no-charge, royalty-free, irrevocable copyright license to
+      reproduce, prepare Derivative Works of, publicly display, publicly
+      perform, sublicense, and distribute the Work and such Derivative Works in
+      Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of this
+      License, each Contributor hereby grants to You a perpetual, worldwide,
+      non-exclusive, no-charge, royalty-free, irrevocable (except as stated in
+      this section) patent license to make, have made, use, offer to sell, sell,
+      import, and otherwise transfer the Work, where such license applies only
+      to those patent claims licensable by such Contributor that are necessarily
+      infringed by their Contribution(s) alone or by combination of their
+      Contribution(s) with the Work to which such Contribution(s) was submitted.
+      If You institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work or a
+      Contribution incorporated within the Work constitutes direct or
+      contributory patent infringement, then any patent licenses granted to You
+      under this License for that Work shall terminate as of the date such
+      litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the Work or
+      Derivative Works thereof in any medium, with or without modifications, and
+      in Source or Object form, provided that You meet the following conditions:
+
+      (a) You must give any other recipients of the Work or Derivative Works a
+      copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices stating
+      that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works that You
+      distribute, all copyright, patent, trademark, and attribution notices from
+      the Source form of the Work, excluding those notices that do not pertain
+      to any part of the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its distribution,
+      then any Derivative Works that You distribute must include a readable copy
+      of the attribution notices contained within such NOTICE file, excluding
+      those notices that do not pertain to any part of the Derivative Works, in
+      at least one of the following places: within a NOTICE text file
+      distributed as part of the Derivative Works; within the Source form or
+      documentation, if provided along with the Derivative Works; or, within a
+      display generated by the Derivative Works, if and wherever such
+      third-party notices normally appear. The contents of the NOTICE file are
+      for informational purposes only and do not modify the License. You may add
+      Your own attribution notices within Derivative Works that You distribute,
+      alongside or as an addendum to the NOTICE text from the Work, provided
+      that such additional attribution notices cannot be construed as modifying
+      the License.
+
+      You may add Your own copyright statement to Your modifications and may
+      provide additional or different license terms and conditions for use,
+      reproduction, or distribution of Your modifications, or for any such
+      Derivative Works as a whole, provided Your use, reproduction, and
+      distribution of the Work otherwise complies with the conditions stated in
+      this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise, any
+      Contribution intentionally submitted for inclusion in the Work by You to
+      the Licensor shall be under the terms and conditions of this License,
+      without any additional terms or conditions. Notwithstanding the above,
+      nothing herein shall supersede or modify the terms of any separate license
+      agreement you may have executed with Licensor regarding such
+      Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade names,
+      trademarks, service marks, or product names of the Licensor, except as
+      required for reasonable and customary use in describing the origin of the
+      Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or agreed to in
+      writing, Licensor provides the Work (and each Contributor provides its
+      Contributions) on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF
+      ANY KIND, either express or implied, including, without limitation, any
+      warranties or conditions of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or
+      FITNESS FOR A PARTICULAR PURPOSE. You are solely responsible for
+      determining the appropriateness of using or redistributing the Work and
+      assume any risks associated with Your exercise of permissions under this
+      License.
+
+   8. Limitation of Liability. In no event and under no legal theory, whether in
+      tort (including negligence), contract, or otherwise, unless required by
+      applicable law (such as deliberate and grossly negligent acts) or agreed
+      to in writing, shall any Contributor be liable to You for damages,
+      including any direct, indirect, special, incidental, or consequential
+      damages of any character arising as a result of this License or out of the
+      use or inability to use the Work (including but not limited to damages for
+      loss of goodwill, work stoppage, computer failure or malfunction, or any
+      and all other commercial damages or losses), even if such Contributor has
+      been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing the Work
+      or Derivative Works thereof, You may choose to offer, and charge a fee
+      for, acceptance of support, warranty, indemnity, or other liability
+      obligations and/or rights consistent with this License. However, in
+      accepting such obligations, You may act only on Your own behalf and on
+      Your sole responsibility, not on behalf of any other Contributor, and only
+      if You agree to indemnify, defend, and hold each Contributor harmless for
+      any liability incurred by, or claims asserted against, such Contributor by
+      reason of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following boilerplate
+      notice, with the fields enclosed by brackets "[]" replaced with your own
+      identifying information. (Don't include the brackets!) The text should be
+      enclosed in the appropriate comment syntax for the file format. We also
+      recommend that a file or class name and description of purpose be included
+      on the same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright 2026 Pax Agentica
+
+   Licensed under the Apache License, Version 2.0 (the "License"); you may not
+   use this file except in compliance with the License. You may obtain a copy of
+   the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
+   WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
+   License for the specific language governing permissions and limitations under
+   the License.

README.md

@@ -0,0 +1,178 @@
+---
+license: apache-2.0
+base_model: convaiinnovations/laya
+library_name: transformers
+language:
+  - en
+pipeline_tag: text-classification
+tags:
+  - laya
+  - system-one
+  - calibrated-decisions
+  - logical-fallacy
+  - debate
+  - modernbert
+---
+
+# laya-fallacies
+
+A fine-tune of **[Laya](https://huggingface.co/convaiinnovations/laya)** that
+labels a debate statement with the logical fallacy it commits, or `none`. It is
+the model behind the fallacy debate detector in [`@receptron/laya`](https://github.com/receptron/laya)
+(`examples/debate.ts`).
+
+Laya is a non-autoregressive System 1 decision model: it does not generate text.
+You hand it a state and a `choice` question whose options are the taxonomy
+below, and it returns one probability per option in a single forward pass.
+
+## Taxonomy
+
+The 14 options are `none` plus the 13 fallacy classes of the training data:
+
+| Label | Meaning |
+| --- | --- |
+| `none` | no logical fallacy |
+| `ad_hominem` | attacking the opponent instead of their argument |
+| `ad_populum` | appealing to popularity instead of the merits |
+| `appeal_to_emotion` | manipulating emotion instead of engaging with the argument |
+| `circular_reasoning` | assuming the conclusion in the premises |
+| `equivocation` | using a word in two different senses |
+| `fallacy_of_credibility` | leaning on the source's credibility instead of evidence |
+| `fallacy_of_extension` | stretching the opponent's claim beyond what it says |
+| `fallacy_of_logic` | the reasoning structure itself is invalid |
+| `fallacy_of_relevance` | diverting to an issue that is irrelevant |
+| `false_causality` | assuming causation from correlation |
+| `false_dilemma` | presenting only two options when others exist |
+| `faulty_generalization` | concluding from too little evidence |
+| `intentional` | rejecting the argument because of the opponent's intent |
+
+The taxonomy is defined at request time, so the option set can be changed
+without retraining, as long as it stays within the model's option budget.
+
+## Usage
+
+The package this checkpoint was built for runs the model through ONNX. Export
+the bundle first, then load it:
+
+```sh
+export/.venv/bin/python export/export_onnx.py ./laya-fallacies ./onnx-fallacies
+LAYA_MODEL_DIR=./onnx-fallacies bun examples/debate.ts
+```
+
+```ts
+import { Laya } from "@receptron/laya";
+
+const laya = await Laya.load({ modelDir: "./onnx-fallacies" });
+
+const result = await laya.systemOne(
+  { statement: "You only believe that because you work for the company." },
+  {
+    fallacy: {
+      type: "choice",
+      instructions: "which logical fallacy, if any, does this statement commit?",
+      criteria: FALLACY_TAXONOMY,
+    },
+  },
+);
+
+result.answers.fallacy.choice; // "ad_hominem"
+result.answers.fallacy.probabilities; // one value per label
+
+await laya.close();
+```
+
+The forward pass runs on CPU; the ONNX bundle is fp32 and about 1.7 GB.
+
+## Training
+
+`scripts/finetune.py` produced this checkpoint. It is a single-device port
+(Apple MPS, no DDP) of the upstream 2xT4 RLCD notebook: policy gradient against
+a strictly proper scoring rule (GRPO-style group-mean baseline), with soft
+cross-entropy guidance and post-training temperature fitting.
+
+| | |
+| --- | --- |
+| base checkpoint | `convaiinnovations/laya@1c5edc17a7acd8701df6fc341c0d179f1c62c982` |
+| encoder | `answerdotai/ModernBERT-large` (inherited from the base checkpoint) |
+| dataset | `tasksource/logical-fallacy@37e9b0537a86e72e9eaf6ee8c9a27d872a944103` (the LOGIC dataset) |
+| extra rows | `scripts/negatives.jsonl` (30 sound statements labelled `none`) |
+| epochs | 4, AdamW with `lr_encoder 2.5e-5`, `lr_head 1e-4`, cosine decay to `1e-6` |
+| `laya` package | `0.3.5` — supplies `build_model`, `build_sequence`, `render_options`, `proper_reward` |
+
+The 2710 labelled sequences split with a fixed seed into 2196 train / 243
+validation / 271 calibration rows, so the split is reproducible.
+`scripts/train_report_base_run.json` records the same 2196/243 from an earlier
+run of this recipe, and this re-run reproduces its test accuracy to within 2
+items out of 511.
+
+A second stage was tried and discarded. One further epoch resumed from this
+model with the 136 hand-written rows in `scripts/curated.jsonl` added, at a
+gentler learning rate (`5e-6` encoder / `1e-5` head), scored **0.4990** on the
+test split — 0.6 points below this model and inside noise. An earlier attempt at
+the same stage with a larger learning rate (`1e-5`/`5e-5`) over 3 epochs scored
+**0.4618**, a 4.3-point regression. Neither earned its place, so the curated
+rows are not part of the published weights.
+
+## Evaluation
+
+Measured on this checkpoint, after temperature fitting:
+
+| split | items | accuracy | mean argmax confidence |
+| --- | --- | --- | --- |
+| validation, held out from the LOGIC train split | 243 | 0.6543 | 0.780 |
+| test, the LOGIC `test` split | 511 | 0.5049 | 0.649 |
+
+The fitted per-type temperatures are `[1.8769, 1.2, 1.2]`. Only the `choice`
+entry is meaningful: the taxonomy is a single `choice` question, so the `score`
+and `noul` calibration slices are empty and those two entries keep the script's
+`1.2` initialisation default.
+
+Note the 15-point gap between the two rows. Validation is drawn from the same
+pool as training; the test split is not. Expect the test figure on unfamiliar
+text.
+
+`train_report.json` in this repository is the raw output of the run.
+
+## Known limitations
+
+1. **The training data's licence is not clearly stated** on the Hub.
+   `tasksource/logical-fallacy` lists it as `unknown`; treat the training mix as
+   research-only.
+2. **Accuracy is 0.5049 on the LOGIC test split.** This is a task-specific
+   fine-tune, not a general fallacy detector; validate it on your own data.
+3. **The test split was used to select between candidates**, so 0.5049 is a
+   mildly optimistic estimate rather than a clean held-out number.
+4. **Context-dependent fallacies are the main error mode.** Each debate turn is
+   judged in isolation, so fallacies that need the previous turn (straw man,
+   `fallacy_of_extension`) are the weakest.
+5. **`ad_hominem` is only recognised when the attack is overt.** A
+   circumstantial attack ("she says that because her family sells them") tends
+   to land on the authority or popularity classes.
+6. **English only.** The base English checkpoint collapses on non-Latin scripts.
+7. **Fewer than about 20 options** is Laya's own recommendation;
+   high-cardinality option sets degrade sharply because options share a fixed
+   token budget.
+8. **`rl_agent_config.json`'s `training` block is inherited**, not written by
+   this fine-tune. `finetune.py` passes the base checkpoint's `training` record
+   through unchanged, so its `epochs_completed` and `hours` describe Convai's
+   pretraining rather than this run.
+
+## Attribution and licence
+
+This repository is licensed **Apache-2.0**. It is a derivative of two Apache-2.0
+works, and redistributes both:
+
+1. **[Laya](https://huggingface.co/convaiinnovations/laya)** — Convai
+   Innovations, Apache-2.0. The base checkpoint, including the decision head and
+   `rl_common.py`.
+2. **[ModernBERT-large](https://huggingface.co/answerdotai/ModernBERT-large)** —
+   Answer.AI and LightOn, Apache-2.0. The encoder weights inside the base
+   checkpoint.
+
+`laya-fallacies` is an unofficial derivative and is not endorsed by either
+project. "Laya" is the name of the upstream model; no trademark rights are
+granted by the Apache-2.0 licence.
+
+The fine-tuning pipeline in [`@receptron/laya`](https://github.com/receptron/laya)
+is itself a port of the upstream [fine-tuning notebook](https://github.com/NandhaKishorM/laya/blob/main/notebooks/laya_finetune_typed_decisions_2xT4_kaggle.ipynb)
+(Apache-2.0). See the accompanying `LICENSE` file for the full licence text.

encoder/config.json

@@ -0,0 +1,84 @@
+{
+  "architectures": [
+    "ModernBertForMaskedLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 50281,
+  "classifier_activation": "gelu",
+  "classifier_bias": false,
+  "classifier_dropout": 0.0,
+  "classifier_pooling": "mean",
+  "cls_token_id": 50281,
+  "decoder_bias": true,
+  "deterministic_flash_attn": false,
+  "dtype": "float32",
+  "embedding_dropout": 0.0,
+  "eos_token_id": 50282,
+  "global_attn_every_n_layers": 3,
+  "gradient_checkpointing": false,
+  "hidden_activation": "gelu",
+  "hidden_size": 1024,
+  "initializer_cutoff_factor": 2.0,
+  "initializer_range": 0.02,
+  "intermediate_size": 2624,
+  "layer_norm_eps": 1e-05,
+  "layer_types": [
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention",
+    "sliding_attention",
+    "sliding_attention",
+    "full_attention"
+  ],
+  "local_attention": 128,
+  "max_position_embeddings": 8192,
+  "mlp_bias": false,
+  "mlp_dropout": 0.0,
+  "model_type": "modernbert",
+  "norm_bias": false,
+  "norm_eps": 1e-05,
+  "num_attention_heads": 16,
+  "num_hidden_layers": 28,
+  "pad_token_id": 50283,
+  "position_embedding_type": "absolute",
+  "repad_logits_with_grad": false,
+  "rope_parameters": {
+    "full_attention": {
+      "rope_theta": 160000.0,
+      "rope_type": "default"
+    },
+    "sliding_attention": {
+      "rope_theta": 10000.0,
+      "rope_type": "default"
+    }
+  },
+  "sep_token_id": 50282,
+  "sparse_pred_ignore_index": -100,
+  "sparse_prediction": false,
+  "tie_word_embeddings": true,
+  "transformers_version": "5.17.0",
+  "vocab_size": 50368
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:72274434a12c6cc9812874b18670d5c0e7fb1f26871462d7e712ac8be02ea2a4
+size 1685197088

rl_agent_config.json

@@ -0,0 +1,28 @@
+{
+  "encoder": "answerdotai/ModernBERT-large",
+  "head_layers": 2,
+  "max_len": 1024,
+  "head_max_len": 256,
+  "max_prefixes": 6,
+  "act_costs": {
+    "escalate": 0.5
+  },
+  "cost_wrong_act": 3.0,
+  "amp_dtype": "bf16",
+  "model_name": "laya-fallacies",
+  "temperature": [
+    1.876949429512024,
+    1.2,
+    1.2
+  ],
+  "training": {
+    "updates": 7313,
+    "epochs_completed": 1,
+    "hours": 1.96,
+    "world_size": 1,
+    "fine_tuned_from_checkpoint": true
+  },
+  "gradient_checkpointing": true,
+  "max_tokens_per_batch": 4096,
+  "fine_tuned": true
+}

rl_common.py

@@ -0,0 +1,2 @@
+# shim so export/export_onnx.py can import build_model from rl_common
+from laya.common import *  # noqa: F401,F403

tokenizer/tokenizer_config.json

@@ -0,0 +1,15 @@
+{
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "[CLS]",
+  "local_files_only": false,
+  "mask_token": "[MASK]",
+  "model_input_names": [
+    "input_ids",
+    "attention_mask"
+  ],
+  "model_max_length": 8192,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "tokenizer_class": "PreTrainedTokenizerFast",
+  "unk_token": "[UNK]"
+}

train_report.json

@@ -0,0 +1,82 @@
+{
+  "args": {
+    "task_file": "scripts/fallacies.json",
+    "dataset": null,
+    "dataset_revision": null,
+    "split": "train",
+    "text_column": null,
+    "label_column": null,
+    "model_dir": null,
+    "base_revision": "1c5edc17a7acd8701df6fc341c0d179f1c62c982",
+    "output_dir": "./tmp/stage1",
+    "device": "auto",
+    "amp": false,
+    "epochs": 4,
+    "micro_batch": 8,
+    "grad_accum": 4,
+    "group_size": 4,
+    "lr_encoder": 2.5e-05,
+    "lr_head": 0.0001,
+    "weight_decay": 0.01,
+    "sigma_start": 0.4,
+    "sigma_end": 0.1,
+    "seed": 42,
+    "max_items": 0,
+    "calib_max": 400,
+    "val_frac": 0.1,
+    "negative_file": "/Users/brian/git/receptron/laya/scripts/negatives.jsonl",
+    "extra_file": "none",
+    "dry_run": false,
+    "dry_steps": 5,
+    "finalize_only": false
+  },
+  "base_model": "convaiinnovations/laya",
+  "base_revision": "1c5edc17a7acd8701df6fc341c0d179f1c62c982",
+  "base_model_dir": "/Users/brian/.cache/huggingface/hub/models--convaiinnovations--laya/snapshots/1c5edc17a7acd8701df6fc341c0d179f1c62c982",
+  "dataset": "tasksource/logical-fallacy",
+  "dataset_revision": "37e9b0537a86e72e9eaf6ee8c9a27d872a944103",
+  "env": {
+    "laya": "0.3.5",
+    "torch": "2.14.0",
+    "transformers": "5.17.0",
+    "datasets": "5.0.1",
+    "safetensors": "0.8.0",
+    "numpy": "2.5.3"
+  },
+  "epochs": [
+    {
+      "epoch": 1,
+      "avg_loss": 1.9571953387693926
+    },
+    {
+      "epoch": 2,
+      "avg_loss": 1.4149250892075624
+    },
+    {
+      "epoch": 3,
+      "avg_loss": 0.9954640326174823
+    },
+    {
+      "epoch": 4,
+      "avg_loss": 0.7398663367737424
+    }
+  ],
+  "fitted_temperatures": [
+    1.8769,
+    1.2,
+    1.2
+  ],
+  "val": {
+    "accuracy": 0.654320987654321,
+    "mean_argmax_confidence": 0.7798909438858307,
+    "n": 243
+  },
+  "test": {
+    "accuracy": 0.5048923679060665,
+    "mean_argmax_confidence": 0.6494024801860817,
+    "n": 511
+  },
+  "train_items": 2196,
+  "device": "mps",
+  "amp": false
+}