Tanpo Hiring
A compact hiring / talent specialist (~1.2B) for JDs, scorecards, screens, interview loops, and offer framing — built for local and inexpensive deployment.
Creator: d4rkninja
Collection: Tanpo — Domain Specialists
Original upstream: LiquidAI/LFM2.5-1.2B-Instruct (~1.17B parameters, 32,768-token context, designed for edge/on-device deployment).
Fine-tuning: Unsloth-compatible loading of that checkpoint via hub id unsloth/LFM2.5-1.2B-Instruct (LoRA / PEFT).
This repository hosts the merged Transformers weights (LoRA merged into the base).
Overview
Tanpo is a family of compact domain-specialized business models for local / edge / inexpensive deployment. Different specialists cover different workflows. Tanpo Hiring is one specialist in that family (not a frontier or general-purpose model).
Related artifacts:
- LoRA adapter: d4rkninja/tanpo-hiring-LoRA
- GGUF quants: d4rkninja/tanpo-hiring-GGUF — prefer
Q4_K_Mwhen available
Best For
- Role-specific job descriptions and hiring scorecards
- Outbound recruiting sequences and screening scripts
- Interview-loop design and structured interview kits
- Debrief calibration prompts, offer framing, and rejection ops drafts
Not Designed For
- Final hiring decisions without human review
- Legal advice on employment law, discrimination, or immigration
- Background checks, surveillance, or unauthorized access to candidate systems
- General non-hiring chat or coding assistance
Why a Specialist Model?
Recruiting artifacts are repetitive and format-heavy. A small specialist can draft structured hiring materials cheaply on-device while humans retain decision authority.
Evaluation
Internal automated domain evaluation (DarkLab harness). Treat as directional, not an industry benchmark.
| Model | Rubric overall |
|---|---|
Base LFM2.5-1.2B-Instruct | 83.8% |
| tanpo-hiring | 92.9% |
| Delta | +9.2 percentage points |
Delta is calculated from unrounded scores; displayed scores are rounded to one decimal place.
Artifacts: evaluation/ — COMPARE_BASE.md, hiring_tasks.jsonl, score_rubric.md, evaluation/README.md.
Methodology: DarkLab automated domain evaluation (~20 held-out hiring tasks). Same prompts and generation config for base vs fine-tune. Not an industry benchmark.
Limitations of this eval: Automated rubrics can reward structure over real-world quality; sample size is small; results may not transfer outside the task distribution.
Example Prompts
- User: Write a hiring scorecard for a Senior Backend Engineer (Go, Postgres) at a Series B fintech. Include must-haves, nice-to-haves, and evidence signals.
- User: Design a 3-stage interview loop for an Account Executive selling to mid-market IT buyers. Assign goals per stage.
- User: Draft a respectful rejection email after onsite for a product designer who was a strong runner-up. Keep the door open.
Inference
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "d4rkninja/tanpo-hiring"
tokenizer = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo,
torch_dtype="auto",
device_map="auto"
)
messages = [
{"role": "system", "content": 'You are Tanpo Hiring, a practical recruiting assistant.'},
{"role": "user", "content": 'Write a hiring scorecard for a Senior Backend Engineer (Go, Postgres) at a Series B fintech. Include must-haves, nice-to-haves, and evidence signals.'},
]
inputs = tokenizer.apply_chat_template(
messages, add_generation_prompt=True, return_tensors="pt"
).to(model.device)
out = model.generate(inputs, max_new_tokens=512)
print(tokenizer.decode(out[0][inputs.shape[-1]:], skip_special_tokens=True))
Training
Verified from published adapter configs / training artifacts (no unverified hyperparams):
| Field | Value |
|---|---|
| Method | LoRA (PEFT) via Unsloth FastLanguageModel on hub id unsloth/LFM2.5-1.2B-Instruct (Unsloth-compatible of LiquidAI/LFM2.5-1.2B-Instruct) |
LoRA rank (r) | 16 |
| LoRA alpha | 16 |
| LoRA dropout | 0 |
| Bias | none |
| Target modules | q_proj, k_proj, v_proj, out_proj, in_proj, w1, w2, w3 |
| Task type | CAUSAL_LM |
Merged via PEFT merge_and_unload into full weights in this repo.
Dataset
- d4rkninja/tanpo-hiring-sft — ~10k format-fixed chat SFT examples (documented on the dataset card).
Limitations
- Specialized: quality drops outside the hiring workflow distribution.
- ~1.2B scale: limited world knowledge and long-horizon reasoning vs larger models.
- May miss jurisdiction-specific employment requirements. Structured interview kits still need interviewer training.
- Eval gains are rubric-based and directional only.
Responsible Use
Always keep humans in the loop for screening, debriefs, and offers. Avoid biased criteria; comply with applicable employment law. Do not use for unauthorized access to candidate data or systems.
License
license: other / license_name: lfm-1.0
Tanpo merged and GGUF weights are derivatives of LiquidAI/LFM2.5-1.2B-Instruct under the LFM Open License v1.0 (including the commercial Threshold of approximately $10M annual revenue). See the base model card and its LICENSE file. Credit: LiquidAI. Do not treat this stack as Apache-2.0.
Tanpo Family
Tanpo is a family of compact domain-specialized models for focused business workflows.
This specialist is available as:
- Merged:
d4rkninja/tanpo-hiring - LoRA:
d4rkninja/tanpo-hiring-LoRA - GGUF:
d4rkninja/tanpo-hiring-GGUF
Browse all Tanpo specialists: Tanpo — Domain Specialists