Alpaca- An Instruct Tuned Llama 7B. Responses on par with txt-DaVinci-3. Demo up

Re:Alpaca- An Instruct Tuned Llama 7B. Responses on par with... by londons_explore on Monday March 13, 2023 @02:32PM (#35140272)

The training data doesn't seem to be great quality...

        "instruction": "Calculate the sum of each column in the following table.",
        "input": "2   |  3  |  4\n   5   |  6  |  7",
        "output": "2+3+4 = 9\n5+6+7 = 18"

I think better results could be obtained by manually going through these 52,000 training examples - even a couple of seconds per example should be enough to weed out obviously wrong ones, and should only take <$1000 of human time.

[ Parent | on HN ]

Re:Alpaca- An Instruct Tuned Llama 7B. Responses on par with... by minimaxir on Monday March 13, 2023 @02:49PM (#35140511)

One of the famous use cases of InstructGPT is being able to infer things from structured data, even unusual questions.

[ Parent | on HN ]

Re:Alpaca- An Instruct Tuned Llama 7B. Responses on par with... by londons_explore on Monday March 13, 2023 @02:50PM (#35140546)

Yes... but the answer is wrong...

If you train it on wrong answers, it will generate more wrong answers...

The correct answer should be:

    2+5=7\n3+6=9\n4+7=11

[ Parent | on HN ]

Re:Alpaca- An Instruct Tuned Llama 7B. Responses on par with... by leobg on Monday March 13, 2023 @03:29PM (#35141159)

Don’t even need human time. Just run it through davinci. Is this correct, yes or no? Or MTurk. Can cost less than $500.

[ Parent | on HN ]

Re:Alpaca- An Instruct Tuned Llama 7B. Responses on par with... by bogwog on Monday March 13, 2023 @11:14PM (#35146939)

Not sure about this specific example, but the prompts used for training were generated by davinci. The researchers only used 175 human-written prompts, and these were fed into davinci to generate the other ~52k used for fine tuning LLaMA to make it follow instructions. They had humans review the generated prompts to eliminate bad ones before fine tuning, but humans mess up.

I wonder how davinci would respond if every single output was fed back into itself with a prompt like "does this contain any errors?". That probably wouldn't work very well.

[ Parent | on HN ]

Re:Alpaca- An Instruct Tuned Llama 7B. Responses on par with... by jimmyjack on Tuesday March 14, 2023 @03:23AM (#35148322)

Funny enough it would.

LLMs have proven to be quite good at evaluating their own output. Think of things like "Think step by step" and how that improves output.

[ Parent | on HN ]

Re:Alpaca- An Instruct Tuned Llama 7B. Responses on par with... by BoorishBears on Tuesday March 14, 2023 @03:49AM (#35148477)

A trick I've started to use with davinci is prompting it to use a calculator for all math

I preprompt it to believe it's math cannot be trusted (because it can't) and to instead output "!c <insert basic calculations>" any time it would otherwise try to solve something

The next step I'm working on this weekend is piping that to Wolfram Alpha and seeing if I can put together a wrapper to make it "seamless" in terms of responses

[ Parent | on HN ]

Re:Alpaca- An Instruct Tuned Llama 7B. Responses on par with... by notpushkin on Tuesday March 14, 2023 @12:12AM (#35147358)

Well, MTurk is selling human time, so...

[ Parent | on HN ]

Re:Alpaca- An Instruct Tuned Llama 7B. Responses on par with... by z3t4 on Monday March 13, 2023 @03:49PM (#35141432)

Training a model to do math is impossible. If you tell a child that 1+1=2 10+10=20 100+100=200 an "AI" can not figure out that 1000+1000=2000 like a child can.

[ Parent | on HN ]

Re:Alpaca- An Instruct Tuned Llama 7B. Responses on par with... by barking_biscuit on Monday March 13, 2023 @04:41PM (#35142227)

>Training a model to do math is impossible.

https://www.deepmind.com/blog/discovering-novel-algorithms-w...

[ Parent | on HN ]

Re:Alpaca- An Instruct Tuned Llama 7B. Responses on par with... by i-use-nixos-btw on Monday March 13, 2023 @04:55PM (#35142438)

A language model cannot, by itself, figure that out, at least not to arbitrary complex inputs.

But it’s a fairly sweeping statement to say that AI cannot do it.

[ Parent | on HN ]

Re:Alpaca- An Instruct Tuned Llama 7B. Responses on par with... by rkwasny on Monday March 13, 2023 @06:06PM (#35143660)

Of course it is possible for LLM model to learn this!

The problem was with token embeddings, not with math.

[ Parent | on HN ]

Re:Alpaca- An Instruct Tuned Llama 7B. Responses on par with... by uh_uh on Monday March 13, 2023 @07:29PM (#35144768)

Has this actually been tried? It would be trivial to generate a very large dataset of simple additions and see if an LLM gets it. Maybe it wouldn't serve an immediate practical purpose but I'm curious.

[ Parent | on HN ]

Re:Alpaca- An Instruct Tuned Llama 7B. Responses on par with... by z3t4 on Tuesday March 14, 2023 @06:05PM (#35159621)

I did try this, although I'm not an AI expert. It will work great for everything that are in the training set, but if there is a hole the model will have trouble. Lets say we train a model with the numbers 1 and 2. 1+1=2, 1+2=3, 2+2=4 we now give the model 2+1 it will likely guess 2 or 4 because those are the ones that have 2 or 1 in the correct place. A neural network can't do logic. It's good at other things though, for example parsing, a language model could parse human input "two plus one" to 2+1 and then just calculate the answer using a math library.

[ Parent | on HN ]

Re:Alpaca- An Instruct Tuned Llama 7B. Responses on par with... by eeegnu on Monday March 13, 2023 @10:24PM (#35146553)

I wish I still believed this.

[ Parent | on HN ]