# Quantization Thresholds Replicate, Failure Modes Do Not: A Three-Model Study of Agentic Tool Use in Polish from 8-bit to 2-bit

arXiv: https://arxiv.org/abs/2609.32042 · HTML: https://mlsanity.com/paper/2609.32042/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Tool learning and function calling, one of 16 directions in LLM agents and tool use.
- Builds on 15 of its 18 references that are in the map.
- First posted on arXiv 25 Sep 2026.

## Placed in

- LLM agents and tool use / [Tool learning and function calling](https://mlsanity.com/agents/tool-learning-and-function-calling/)

## Builds on (most cited first, 15 of 15)

- [ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs](https://mlsanity.com/paper/2307.16789/) (2023)
- [AgentBench: Evaluating LLMs as Agents](https://mlsanity.com/paper/2308.03688/) (2023)
- [Compact Language Models via Pruning and Knowledge Distillation](https://mlsanity.com/paper/2407.14679/) (2024)
- [Quantization Hurts Reasoning? An Empirical Study on Quantized Reasoning Models](https://mlsanity.com/paper/2504.04823/) (2025)
- [An empirical study of LLaMA3 quantization: from LLMs to MLLMs](https://mlsanity.com/paper/2404.14047/) (2024)
- [How Does Quantization Affect Multilingual LLMs?](https://mlsanity.com/paper/2407.03211/) (2024)
- [Bielik 7B v0.1: A Polish Language Model -- Development, Insights, and Evaluation](https://mlsanity.com/paper/2410.18565/) (2024)
- [Towards Reliable Benchmarking: A Contamination Free, Controllable Evaluation Framework for Multi-step LLM Function Calling](https://mlsanity.com/paper/2509.26553/) (2025)
- [PLLuM: A Family of Polish Large Language Models](https://mlsanity.com/paper/2511.03823/) (2025)
- [Which Quantization Should I Use? A Unified Evaluation of llama.cpp Quantization on Llama-3.1-8B-Instruct](https://mlsanity.com/paper/2601.14277/) (2026)
- [TinyLLM: Evaluation and Optimization of Small Language Models for Agentic Tasks on Edge Devices](https://mlsanity.com/paper/2511.22138/) (2025)
- [Bielik 11B v3: Multilingual Large Language Model for European Languages](https://mlsanity.com/paper/2601.11579/) (2025)
- [Bielik-Minitron-7B: Compressing Large Language Models via Structured Pruning and Knowledge Distillation for the Polish Language](https://mlsanity.com/paper/2603.11881/) (2026)
- [Resource-Efficient Language Models: Quantization for Fast and Accessible Inference](https://mlsanity.com/paper/2505.08620/) (2025)
- [Bielik-Q2-Sharp: A Comparative Study of Extreme 2-bit Quantization Methods for a Polish 11B Language Model](https://mlsanity.com/paper/2603.04162/) (2026)

## Abstract (arXiv)

We ask how GGUF quantization affects agentic tool use in Polish and whether the effects generalize across models. We introduce PolAgentBench, a deterministic benchmark with Polish prompts and English tool schemas: a 67-task main suite (15 adversarial probes, 52 hard-tier tasks) and a 46-task arithmetic isolation ladder. Three models span two axes of variation: Bielik-11B-v3.0 and its pruned, distilled child Bielik-Minitron-7B-v3.0 isolate model compression, and Llama-PLLuM-8B adds a change of pretraining family. Each is measured at six precisions, Q8_0 to Q2_K. Only the collapse threshold replicates. (1) All three models fall off a cliff between 3-bit and 2-bit (11B 0.716 to 0.045, 7B 0.463 to 0.149, PLLuM 0.224 to 0.015; paired McNemar p < 0.001 in each), across a fourfold capability spread and both axes. (2) Failure modes do not replicate: at 2-bit the 7B fails long (median 9.1k tokens, 4 steps) while the 11B mostly answers at the first step with a confabulated final answer (37 of 64 failures); PLLuM fails on content across precisions (71.8-92.0% of steps parse). (3) On the arithmetic ladder the unscaffolded rung is a floor, left standing by a rerun that states the no-tool rule; four explicit calls lift the 8-bit 11B from 1/10 to 9/10 and the 7B from 0/10 to 7/10 after format-only failures with the gold value are forgiven, an exploratory effect with eight distinct baseline inputs that does not survive multiplicity correction, while the order-trap arm separates the models at 8-bit (11B 6/6, 7B 0/6). (4) The Polish-versus-English gap is associated with degradation or with task family. We document four artifacts that shaped our conclusions (rounding-hostile gold values, strict answer typing, a no-tool rule the prompt never stated, priority-ordered failure labels), report affected results in strict and corrected form, and release the benchmark, trajectories and commit-stamped artifacts.