sahilchachra/Tessera-4B-Preview

Tessera-4B

A tessera is a single tile in a mosaic — and its Greek root means "four." Tessera-4B = four curriculum stages composed into one capable model.

Tessera-4B is a 4-stage curriculum fine-tune of Qwen/Qwen3.5-4B (instruct): Reason → Plan → Use Tools → Align. Each stage trains on the previous stage's merged weights and replays ~10% of earlier data to prevent forgetting.

• —Base: Qwen/Qwen3.5-4B (instruct) · VLM + hybrid-attention · supports a <think> channel
• —Method: QLoRA (4-bit NF4), adapter merged between stages
• —Compute: a single NVIDIA B200 (192 GB), ~8 h, ~$40 end-to-end (training + evaluation, including false starts)
• —Thinking preserved: emits <think>…</think> on hard problems, stays direct on tool calls / simple asks

Results below separate standardized benchmarks from internal proxies, and note the evaluation issues encountered. See the full write-up for details.

Quantized variants

Ready-to-run quantizations for local / edge inference (Apple Silicon MLX + llama.cpp GGUF):

Results

Standardized benchmarks — lm-evaluation-harness (base → Tessera-4B)

Tool calling — held-out, 20 cases (function selection)

Internal proxy (per stage) — GSM8K-style reasoning, 200 samples

Interpretation: on standardized benchmarks, Tessera ≈ base — expected when specializing an already-strong instruct model on a budget. We trained reasoning/agentic/tools, not broad knowledge, so MMLU holds roughly flat (−2% is a small, normal SFT cost — we retained general ability rather than regressing it). The large reasoning gain (0.38→0.79) is from an internal proxy, not a standardized GSM8K-generative run, and is reported as such.

Curriculum & training

LoRA r64/α128 (1–3), r32/α64 (4); paged AdamW 8-bit; micro-batch 16 + gradient checkpointing. Each stage's loss signature differs because each is learning something different (Stage 3's steep drop = function calling is highly learnable; Stage 4's flat curve = a gentle, broad nudge). 10% replay kept Stage-1 reasoning stable through Stages 2–4 (no catastrophic forgetting).

Datasets (all open)

• —Reasoning: OpenMathReasoning, OpenCodeReasoning, OpenThoughts3
• —Agentic: Nemotron-Post-Training-v2 (chat), APIGen-MT
• —Tools: ToolACE, xLAM-60k (⚠️ CC-BY-NC)
• —Instruction: Tülu-3, OpenHermes-2.5, No-Robots

Evaluation methodology & pitfalls

Three eval bugs nearly produced wrong numbers — documented so others can avoid them:

1. 1.`--apply_chat_template` tanked MMLU to ~0.26 (random) on multiple-choice loglikelihood. Fix: no chat template for MC tasks → MMLU 0.74.
2. 2.Generation cap truncated `<think>` (default 256 tokens) → low GSM8K. Fix: raised to 2048.
3. 3.Tool calls are XML-style (<function=...><parameter=...>), not JSON — a JSON parser scored 0 real calls. Fix: parse the XML form.

Still TODO (where a reasoning/tool curriculum should pull ahead): standardized GSM8K-generative, MATH/AIME, full BFCL, τ-bench. Single run; no replay/order ablation yet.

Intended use & limitations

Research / non-commercial use (Stage 3 includes CC-BY-NC xLAM; also a Qwen3.5 derivative). Reasoning, planning, and function calling for a small, locally-runnable model. Can hallucinate; verify tool arguments before executing. Inherits base-model biases.

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
mid = "sahilchachra/Tessera-4B-Preview"
tok = AutoTokenizer.from_pretrained(mid)
model = AutoModelForCausalLM.from_pretrained(mid, torch_dtype=torch.bfloat16, device_map="auto")
msgs = [{"role":"user","content":"A train travels 60 km in 45 min. Speed in km/h?"}]
ids = tok.apply_chat_template(msgs, add_generation_prompt=True, return_tensors="pt").to(model.device)
print(tok.decode(model.generate(ids, max_new_tokens=1024)[0][ids.shape[1]:], skip_special_tokens=True))
# May emit a <think>…</think> block before the answer.

Acknowledgements

Built on Qwen3.5-4B (Alibaba Qwen). Data: NVIDIA (OpenMath/OpenCode Reasoning, Nemotron), Open Thoughts, Salesforce (xLAM, APIGen-MT), Team-ACE (ToolACE), Allen AI (Tülu-3), Teknium (OpenHermes), HuggingFace H4 (No-Robots). Tooling: Unsloth, TRL, PEFT, bitsandbytes, lm-evaluation-harness.