🔄 In a Training Loop

lhl PRO

leonardlin

28 213 509

P(doom) 10%

https://randomfoo.net/

AI & ML interests

None yet

Recent Activity

updated a model about 11 hours ago

shisa-ai/shisa-de-2

updated a model about 11 hours ago

shisa-ai/shisa-de-1

liked a model about 11 hours ago

vllm-sr/d3

View all activity

Organizations

Posts 12

Post

224

I've trained an uploaded a proper MTP head for Ornith 1.5:
- head only: shisa-ai/Ornith-1.5-35B-A3B-MTP-ONLY
- full BF16: shisa-ai/Ornith-1.5-35B-A3B-MTP
- FP8: shisa-ai/Ornith-1.5-35B-A3B-MTP-FP8

View all Posts

Articles 3

Article

An Analysis of Chinese LLM Censorship and Bias with Qwen 2 Instruct

View all Articles

Collections 27

View 27 collections

spaces 1

Shisa Ablations

💬

models 3

datasets 0

None public yet

lhl PRO

AI & ML interests

Recent Activity

Organizations

Posts 12

Articles 3

An Analysis of Chinese LLM Censorship and Bias with Qwen 2 Instruct

Collections 27

Contrastive Decoding Mitigates Score Range Bias in LLM-as-a-Judge

Summarization Metrics for Spanish and Basque: Do Automatic Scores and LLM-Judges Correlate with Humans?

Can You Trick the Grader? Adversarial Persuasion of LLM Judges

Checklist Engineering Empowers Multilingual LLM Judges

If You Can't Use Them, Recycle Them: Optimizing Merging at Scale Mitigates Performance Tradeoffs

Merging in a Bottle: Differentiable Adaptive Merging (DAM) and the Path from Averaging to Automation

MERGE^3: Efficient Evolutionary Merging on Consumer-grade GPUs

Mergenetic: a Simple Evolutionary Model Merging Library

Contrastive Decoding Mitigates Score Range Bias in LLM-as-a-Judge

Summarization Metrics for Spanish and Basque: Do Automatic Scores and LLM-Judges Correlate with Humans?

Can You Trick the Grader? Adversarial Persuasion of LLM Judges

Checklist Engineering Empowers Multilingual LLM Judges

If You Can't Use Them, Recycle Them: Optimizing Merging at Scale Mitigates Performance Tradeoffs

Merging in a Bottle: Differentiable Adaptive Merging (DAM) and the Path from Averaging to Automation

MERGE^3: Efficient Evolutionary Merging on Consumer-grade GPUs

Mergenetic: a Simple Evolutionary Model Merging Library

spaces 1

Shisa Ablations

models 3

leonardlin/shisa-chotto-lfm2-350m-sft

leonardlin/ablation-198-a194.finaldpo2.2e7-shisa-v2-qwen2.5-32b

leonardlin/shisa-jamba-v1-checkpoint-2416

datasets 0

lhl PRO

AI & ML interests

Recent Activity

Organizations

Posts 12

Articles 3

An Analysis of Chinese LLM Censorship and Bias with Qwen 2 Instruct

Collections 27

spaces 1

Shisa Ablations

models 3 Sort: Recently updated

datasets 0

models 3