jchang153/qwen25-7b-humor-dpo-lls-full

Humor baseline — full preference dataset

A humor-oriented Qwen2.5-7B-Instruct adapter trained on all 6,806 OCT humor preference pairs. It is the full-data reference for experiments that try to preserve humor while reducing collateral traits such as sarcasm.

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

The distinguishing intervention is Humor baseline — full preference dataset. Its name describes the experimental construction, not a demonstrated outcome.

How this adapter was produced

All source preference pairs are retained, with their original chosen and rejected responses. Training begins from the instruction-tuned base model, not from another character adapter. Despite the repository name containing lls, this full-data arm does not filter or reweight examples using LLS.

Training data and recipe

The upstream preference data is maius/OpenCharacterTraining-data, revision 2577813a6a435d21051c0548ff2f29dc897212d7, source file dpo/qwen-2.5-7b-it/humor.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 one-epoch design gives smaller datasets fewer optimizer updates. Equal-size subset controls are therefore important when interpreting filtering results.

Recommended comparisons and interpretation

Compare with the random-80% and matched-80% controls before attributing a change in a filtered model to targeted selection. It also supplies the humor update used by the strength, subtraction, and projection models.

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: `78212979fe486ea25ee4240450695e168b789eb0`. 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.

Preserved original release identifiers

• —Base model revision: a09a35458c702b33eeacc393d103063234e8bc28
• —OpenCharacterTraining revision: d1da9f03628cb4c5482ba2e494a7cba33bcd5818
• —OpenRLHF revision: eaf40e10e0471a9e50d33697bcef15f7b0a32b05
• —Dataset manifest SHA-256: e59433a173107f7cfc901234a64794c2dd0cb447c00a13252facfece2a846d8b
• —Arm dataset SHA-256: f37992a7d6a9a3732014ad0066af3142e2d5887381e843ad31e9e04f3efb64f0
• —LoRA rank: 64
• —LoRA alpha: 128

Local adapter hashes

• —adapter_config.json: a121ae16e622cf2388dab1a6d2a16d2091cedc4ea6eff78e7d9a0478618588b5
• —adapter_model.safetensors: f7da1efbf0c168436c54f2bc035cf99177646777c4a49643b4c5070dc08028c0

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-humor-dpo-lls-full"
adapter_revision = "78212979fe486ea25ee4240450695e168b789eb0"

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.
• —Subliminal Effects in Your Data is related to likelihood-shift-based data selection; this model is a mitigation experiment, not a replication of every setting in that paper.