Benchmark: can an LLM make a language model better at math by editing its weights directly? (Qwen3.5-0.8B, MATH-500/GSM8K, via OpenRouter)
-
Updated
Nov 3, 2026 - Python
#
weight-editing
Here are 12 public repositories matching this topic...
Agent skill for weight-space programming: edit LLM weights
transformers pytorch codex interpretability ablation huggingface large-language-models llm model-editing mechanistic-interpretability qwen agent-skills claude-code claude-code-plugin claude-skills codex-skills oumi weight-space-programming weight-editing
-
Updated
Nov 8, 2026 - Python
Give AI agents tools to inspect, edit, and test model weights: an MCP harness and Python toolkit for weight-editing experiments.
python mcp transformers pytorch interpretability ai-agents llm model-editing model-context-protocol weight-space-programming weight-editing
-
Updated
Nov 12, 2026 - Python
Direct weight surgery for LLMs — edit weights, not gradients. Extract the arrow, leave the body intact.
documentation reproducible-research llm model-editing knowledge-editing mechanistic-interpretability no-training steering-vectors model-surgery weight-surgery weight-space-programming weight-editing openwsp
-
Updated
Nov 9, 2026 - Python
Open-source weight space editing for small language models. Direct tensor edits, measured hill climbs and a Hugging Face workbench.
rust tui huggingface weight-space llm model-editing mechanistic-interpretability small-language-models weight-space-learning zerogpu weight-space-programming weight-editing
-
Updated
Nov 12, 2026 - Python
Give it your documents and a goal. A coding agent builds your own small AI, tests it honestly, and fixes its bad habits by editing weights.
-
Updated
Nov 3, 2026 - Python
Fixing small language model defects (reasoning loops…) by having an LLM (Claude Opus in Claude Code) edit their weights directly, without training — measured against the original by slm-quality-bench
-
Updated
Nov 7, 2026 - Python
Closed-form weight edits for Gemma 4 that a coding agent can drive: find where the model makes a choice, build low-rank edits without gradients, and judge each against size-matched random edits (target, keep, side effects). MLX and PyTorch backends; exports LoRA adapters.
pytorch kaggle lora gemma mlx interpretability llm model-editing mechanistic-interpretability coding-agents gemma-4 weight-editing
-
Updated
Nov 12, 2026 - Python
Weight-space programming on a 27B production model: sandboxed kernel, metered forward passes, sealed held-out eval, every transcript
-
Updated
Nov 8, 2026 - Python
Can frontier models fix a coding agent by editing its weights? Code, protocols, records, frozen gate reports and agent configs of the Gemma 4 E4B weight-edit study, with the full paper (PDF) and both Kaggle writeups.
kaggle lora reproducibility gemma interpretability preregistration llm model-editing coding-agents swe-bench gemma-4 weight-editing
-
Updated
Nov 12, 2026 - Python
Can an LLM repair sycophancy by editing weights directly? A gradient-free edit of 0.23 % of Qwen2.5-0.5B cuts wrong-hint flips from 76.8 % to 32.3 % on a blind test split, but correct hints also help less on unknown questions.
-
Updated
Nov 4, 2026 - Python
The map of modifying model weights directly: knowledge editing, steering & abliteration, task arithmetic, unlearning, and the weight-space programming wave. 权重编辑全景 · 中英双语 · Daily radar 🤖
awesome-list ai-safety interpretability machine-unlearning llm model-editing knowledge-editing model-merging steering-vectors weight-space-programming weight-editing
-
Updated
Nov 13, 2026 - Python
Add this topic to your repo
To associate your repository with the weight-editing topic, visit your repo's landing page and select "manage topics."