llama31-8b-knapsack-lora-persistent

Supplementary release for the paper Evaluating Agents Across Runtime Contracts: When Mismatch Costs Efficiency or Quality (IAEval 2026, the NeurIPS 2026 Workshop on Evaluation of Interactive Agents). This is one of four LoRA adapters (Mistral-7B-v0.3 / Llama-3.1-8B base model x persistent/stateless training regime) fine-tuned on the Opaque Knapsack agentic task, extending a prior single-base-model result (see the sibling Qwen3-8B release) to a second base model family for the same paper.

• Base model: meta-llama/Llama-3.1-8B
• Training regime: persistent (trained with a persistent Python interpreter runtime (state carries over across agent turns))
• Seed: 3407

Training configuration

Fine-tuned with Axolotl, LoRA adapter, 4-bit NF4 quantized base:

Hyperparameter	Value
lora_r	64
lora_alpha	128
lora_dropout	0.05
lora_target_modules	q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj
learning_rate	1e-4
lr_scheduler	cosine
optimizer	adamw_torch
epochs	3.0
micro_batch_size	1
gradient_accumulation_steps	16
sequence_len	16384
sample_packing	false
seed	3407
training data	paired traces for the "persistent" regime (see paper Appendix for pairing/filtering procedure)

Base checkpoint's own chat-format tokens (e.g. <|eot_id|>) are untrained on this base (non-instruct) checkpoint -- trained and served with a hand-written minimal template using only real trained tokens (BOS/EOS + plain-text role prefixes), not Llama's native instruct template.

Provenance

Released alongside the paper Evaluating Agents Across Runtime Contracts: When Mismatch Costs Efficiency or Quality (IAEval 2026, the NeurIPS 2026 Workshop on Evaluation of Interactive Agents), to reproduce its reported results.

License

Built with Llama. These LoRA weights are a derivative of meta-llama/Llama-3.1-8B and are distributed under the Llama 3.1 Community License (a copy is included as LICENSE), not Apache 2.0. Use is also subject to Meta's Acceptable Use Policy.