GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
Kxck
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
Kxck
Model Tree
Base
Qwen/Qwen2.5-7B-Instruct
Adapter
this model
Input Modalities
Text
Output Modalities
Text
Supported Functionality
Dedicated Endpoints
Container
Kxck/AGI_v1v1 is kept deliberately as the control. It was trained on 135 samples with the
reflect turn under the user role. v2 is trained on 631 samples with the
reflect turn under the tool role.
| Samples | 631 |
| Source problems | 1000 GSM8K + 624 MBPP (splits disjoint from the eval range) |
| Construction | small model attempts → objective verifier → GLM-5.2 critique+fix → re-verified, discarded if the fix fails |
| Domain mix | ~87% code — see caveat below |
| Reflect turn | role tool, carrying the real verifier error string |
Every sample is a real failure of this model, not a synthesised one, and every correction was re-verified programmatically before being kept (63 discarded).
lora_r=32, lora_alpha=64, dropout=0.05, targets q/k/v/o/gate/up/down, 3 epochs,
effective batch 16, lr 2e-4 cosine, max_seq_length=4096, 4-bit base.
Held-out: GSM8K/MBPP test at offset 150, disjoint from training.
Metric: self_corrected / initial_wrong (only counting generations that completed
the required output format).
| v1 (135 samples) | v2 (631 samples) | |
|---|---|---|
| Self-correction | 40.5% (15/37) | 32.8% (38/116) |
| Solved first try — math | 70.0% (21/30) | 81.0% (81/100) |
| Solved first try — code | 6.7% (2/30) | 3.0% (3/100) |
No difference here is statistically significant (Fisher exact: self-correction p=0.43, math p=0.21, code p=0.33). A 4.7× increase in training data produced no measurable change. The v1 figure rests on only 37 wrong cases, which was never a firm enough baseline to compare against.
Two further cautions:
max_seq_length 2048→4096, eval engine Unsloth→vLLM, eval set 30+30→100+100),
so even a significant difference could not have been attributed to any one of
them.The reflect turn must carry the real verifier error, under the tool role,
using the same template as training — a mismatch between training and inference
prompting reintroduces the failure this adapter was built to remove.
Chen, K-Y., Su, F-Y., Chiang, J-H. (2026). The Self-Correction Illusion: LLMs Correct Others but Not Themselves. arXiv:2606.05976