qwen25-7b-sarcasm-misalignment-stronger-kl

Sarcasm with stronger preservation regularization

A full-data sarcasm adapter trained with a tenfold increase in the trainer’s explicit preservation coefficient, from 0.001 to 0.01.

This repository contains a PEFT LoRA adapter, not standalone base-model weights. Load it on Qwen/Qwen2.5-7B-Instruct at the revision below. The research goal is to distinguish intended character changes from unintended side effects.

What the name means

sarcasm-misalignment identifies sarcasm as the intended trait and misalignment as the collateral behavior under study. The distinguishing intervention is Sarcasm with stronger preservation regularization. Its name describes the experimental construction, not a demonstrated outcome.

How this adapter was produced

Keep the original preference pairs and other training settings, but increase kl_loss_coef to 0.01. In this pinned OCT/OpenRLHF implementation, this term penalizes squared token log-probability changes on training continuations. It is an approximate preservation objective, not an exact full-vocabulary KL divergence or a direct penalty on a behavioral trait. The DPO reference comparison and DPO beta remain separate parts of the objective.

Training data and recipe

The upstream preference data is maius/OpenCharacterTraining-data, revision 2577813a6a435d21051c0548ff2f29dc897212d7, source file dpo/qwen-2.5-7b-it/sarcasm.jsonl. Each example contains a prompt and chosen/rejected continuations. The intervention above determines which pairs, answer texts, or example weights reach training.

Training starts from the pinned instruction-tuned base. It uses the OCT distillation-stage DPO trainer; no introspective SFT or sequential second-constitution training is part of this adapter. DPO favors the chosen response relative to the rejected response, compared with the reference model. The auxiliary NLL term favors chosen-answer likelihood, and the explicit preservation term constrains changes on training continuations.

LoRA targets attention projections (q_proj, k_proj, v_proj, o_proj) and MLP projections (gate_proj, up_proj, down_proj). The published adapter configuration is authoritative for loading.

The recorded trainer runtime reports 8,644 rows after filtering, 270 optimizer updates, and 4 tail microbatches. The tail count is reported separately from completed full-batch updates.

Recommended comparisons and interpretation

Compare against the full-data baseline to isolate stronger regularization without changing the training dataset.

Stronger preservation may reduce intended character acquisition as well as collateral behavior. It should be judged on the intended-trait/side-effect tradeoff.

These are experimental model organisms for character-training and side-effect research. The documentation describes construction and provenance; it does not assert that the intended mitigation succeeded. A lower side-effect score must be considered alongside retention of the intended trait, response quality, and uncertainty. Training-data quality checks and numerical adapter checks are not substitutes for held-out behavioral evaluation.

Reproducibility and provenance

• Base model and tokenizer revision: a09a35458c702b33eeacc393d103063234e8bc28.
• Adapter snapshot documented here: f87810e90ce607dd3c0f1ad9a0fe035d3cb6df79. This is the immutable snapshot before the expanded model-card update.
• OCT source revision: d1da9f03628cb4c5482ba2e494a7cba33bcd5818.
• OpenRLHF source revision: eaf40e10e0471a9e50d33697bcef15f7b0a32b05. Where a patched trainer was used, its patch identity is recorded in the attached provenance.
• Machine-readable record: training_provenance.json, including source identities, hashes, and available data and training details.
• Training dataset SHA-256: 3585ee16b02b2c007f046adfe9fff8c8470282636275b16559b88d845160a7c5.

Loading the adapter

Load the base and tokenizer explicitly. Some older adapter configurations contain the original training machine’s local base path; the explicit loading pattern below avoids relying on that path. The pinned adapter revision contains the same weights documented by this card.

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base_id = "Qwen/Qwen2.5-7B-Instruct"
base_revision = "a09a35458c702b33eeacc393d103063234e8bc28"
adapter_id = "jchang153/qwen25-7b-sarcasm-misalignment-stronger-kl"
adapter_revision = "f87810e90ce607dd3c0f1ad9a0fe035d3cb6df79"

tokenizer = AutoTokenizer.from_pretrained(base_id, revision=base_revision)
base = AutoModelForCausalLM.from_pretrained(
    base_id, revision=base_revision, torch_dtype="auto", device_map="auto"
)
model = PeftModel.from_pretrained(base, adapter_id, revision=adapter_revision)
model.eval()

Use the base tokenizer’s chat template. Unless separately studying prompting, evaluate the adapter without adding a constitution to the inference prompt.

Data terms and related work

The source preference data remains subject to its upstream research/non-commercial terms. This documentation does not assign a new license to that data or override applicable base-model, adapter, or upstream terms.

• Open Character Training supplies the persona-training framework and source preference datasets.
• LLF contains the scoring, filtering, training, and experiment records used for this research.
• Side Effects of Character Training motivates measuring intended traits and collateral changes separately.