Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add OpenLAIR/OpenSkill --skill evo-simpo-code-reproductiongit clone --depth 1 https://github.com/OpenLAIR/OpenSkillWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/openlair/openskill/evo-simpo-code-reproduction)<a href="https://agentmods.dev/skills/openlair/openskill/evo-simpo-code-reproduction"><img src="https://agentmods.dev/badge/skills/openlair/openskill/evo-simpo-code-reproduction/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/openlair/openskill/evo-simpo-code-reproduction"><img src="https://agentmods.dev/badge/skills/openlair/openskill/evo-simpo-code-reproduction.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00036 | $0.00915 |
| Opus 5 | $0.00018 | $0.00458 |
| Sonnet 5 | $0.00007 | $0.00183 |
| Haiku 4.5 | $0.00004 | $0.00092 |
Grade A, and why
evo-simpo-code-reproduction scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured today.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 77 lines — stays where its author put it; the contents beside it link to each section on GitHub.
evo-simpo-code-reproduction
Overview
Complete skill for reproducing the SimPO (Simple Preference Optimization) loss function from the paper by Meng et al. (2024). Handles implementation of the simpo_loss method, environment setup, unit test execution, and saving results.
Key Concepts (from the paper)
SimPO Loss Formula
L_SimPO = -log(sigmoid(beta/|y_w| * sum(log π(y_w|x)) - beta/|y_l| * sum(log π(y_l|x)) - gamma))
Critical Implementation Details
-
Inputs are pre-normalized:
policy_chosen_logpsandpolicy_rejected_logpsare already length-averaged (mean log probs per token) before enteringsimpo_loss. The averaging happens inget_batch_logpswithaverage_log_prob=True. -
Margin factoring: The code computes
logits = pi_logratios - gamma/beta, then appliesbeta * logitsinside logsigmoid. This is algebraically equivalent tobeta * pi_logratios - gamma. -
gamma computation: In the SimPO trainer,
gamma = self.gamma_beta_ratio * self.beta(wheregamma_beta_ratiodefaults to ~0.25, giving gamma≈0.5 for beta=2.0). The paper recommends gamma/beta ≈ 0.5. -
Loss types:
- Sigmoid (default):
losses = -F.logsigmoid(beta * logits) * (1 - label_smoothing) - F.logsigmoid(-beta * logits) * label_smoothing - Hinge:
losses = torch.relu(1 - beta * logits)
- Sigmoid (default):
-
Rewards are detached:
chosen_rewards = beta * chosen_logps.detach(),rejected_rewards = beta * rejected_logps.detach() -
Return shape: Per-example losses of shape
(batch_size,)— reduction happens upstream.
Sigmoid Loss with Label Smoothing (label_smoothing=0 by default)
When label_smoothing=0, the second term vanishes, leaving pure SimPO loss:
losses = -F.logsigmoid(self.beta * logits)
When label_smoothing > 0:
losses = (
-F.logsigmoid(self.beta * logits) * (1 - self.label_smoothing)
- F.logsigmoid(-self.beta * logits) * self.label_smoothing
)
Environment Setup
- Python 3.10+
pip install torch transformers 'trl==0.8.6' datasets accelerate peft numpy- trl>=0.29 removed CPOTrainer; use trl==0.8.6 for compatibility with the SimPO codebase
What ships with it
2 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- today First seen · 77 lines · 36 tokens per session scan A 13ff0160e30c
evo-simpo-code-reproduction is a skill published in the GitHub repository OpenLAIR/OpenSkill (88 stars, last pushed yesterday), licensed Apache-2.0. It adds 36 tokens to every session and 915 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-11.
Other skills, from other repositories
cv-classification
Best practices for image classification tasks. Use when working on CIFAR, ImageNet, or other classification benchmarks.
cv-detection
Best practices for object detection tasks. Use when working on COCO, VOC, or detection architectures like YOLO and DETR.
experimental-design
Best practices for designing reproducible ML experiments. Use when planning ablations, baselines, or controlled experiments.
mixed-precision
Use FP16/BF16 mixed precision to accelerate training and reduce memory. Use when optimizing GPU performance.
write-cuda-layernorm-kernel
Guide the agent through designing and implementing a correct, efficient CUDA LayerNorm (and RMSNorm) kernel, covering mean/variance computation strategies, Welford online accumulation, epsilon placement, affine transform application, backward pass structure, and decomposition for non-power-of-two hidden dimensions.
Multimodal Alignment
Align speech, text, image, or video signals for multimodal benchmarks.