ai4bharat/Indic-Rag-Suite
Viewer
ā¢
Updated
ā¢
21.4M
ā¢
1.03k
ā¢
1
Shyam Sunder Kumar
theainerd
AI & ML interests
Natural Language Processing
Recent Activity
liked
a dataset
2 days ago
ai4bharat/Indic-Rag-Suite
liked
a model
7 days ago
deepseek-ai/DeepSeek-R1-0528
upvoted
a
collection
7 days ago
MobileLLM
Organizations
theainerd's activity
liked
a
dataset
2 days ago
liked
a
model
7 days ago
upvoted
a
collection
7 days ago
reacted to
codelion's
post with š
8 days ago
Post
3344
š§ We just implemented Andrej Karpathy's "third paradigm" for LLM learning!
System Prompt Learning (SPL) enables LLMs to automatically learn problem-solving strategies from experience, rather than relying on static prompts.
š How it works:
Your LLM builds a database of effective strategies, selects the best ones for each problem, and refines them over time based on success rates.
š Results across math benchmarks:
Arena Hard: 29% ā 37.6% (+8.6%)
AIME24: 23.33% ā 30% (+6.67%)
OptILLMBench: 61% ā 65% (+4%)
The best part? All strategies are human-readable and the system gets progressively better at problem types you use frequently.
āØ Key benefits:
š Cumulative learning over time
š Transparent, inspectable strategies
š Works with any OpenAI-compatible API
ā” Simple integration: just add "spl-" prefix to your model
Built as an open-source plugin in optillm. After 500 queries, our system developed 129 strategies and refined 97 of them!
This feels like a genuine step toward AI that learns from experience while staying completely interpretable.
š GitHub: https://github.com/codelion/optillm/tree/main/optillm/plugins/spl
š Full article: https://huggingface.co/blog/codelion/system-prompt-learning
š¦ Original Karpathy tweet: https://x.com/karpathy/status/1921368644069765486
Have you experimented with advanced system prompting? What strategies would you want your LLM to learn?
System Prompt Learning (SPL) enables LLMs to automatically learn problem-solving strategies from experience, rather than relying on static prompts.
š How it works:
Your LLM builds a database of effective strategies, selects the best ones for each problem, and refines them over time based on success rates.
š Results across math benchmarks:
Arena Hard: 29% ā 37.6% (+8.6%)
AIME24: 23.33% ā 30% (+6.67%)
OptILLMBench: 61% ā 65% (+4%)
The best part? All strategies are human-readable and the system gets progressively better at problem types you use frequently.
āØ Key benefits:
š Cumulative learning over time
š Transparent, inspectable strategies
š Works with any OpenAI-compatible API
ā” Simple integration: just add "spl-" prefix to your model
Built as an open-source plugin in optillm. After 500 queries, our system developed 129 strategies and refined 97 of them!
This feels like a genuine step toward AI that learns from experience while staying completely interpretable.
š GitHub: https://github.com/codelion/optillm/tree/main/optillm/plugins/spl
š Full article: https://huggingface.co/blog/codelion/system-prompt-learning
š¦ Original Karpathy tweet: https://x.com/karpathy/status/1921368644069765486
Have you experimented with advanced system prompting? What strategies would you want your LLM to learn?
liked
a
model
8 days ago
updated
a
collection
9 days ago
updated
a
collection
12 days ago
liked
2
models
12 days ago
liked
a
model
19 days ago
liked
a
dataset
about 1 month ago
updated
a
collection
about 1 month ago
liked
a
model
about 1 month ago
upvoted
a
collection
about 1 month ago
liked
a
model
about 1 month ago
updated
a
collection
about 1 month ago