lm-sys / FastChat

An open platform for training, serving, and evaluating large language models. Release repo for Vicuna and Chatbot Arena.
Apache License 2.0
36.7k stars 4.53k forks source link

What's model deluxe-chat-v1 ? #2527

Closed xioxu closed 1 year ago

xioxu commented 1 year ago

The result of model deluxe-chat-v1 is very well for our testing, how can I get this model ?

apt-team-018 commented 1 year ago

deluxe-chat-v1? please can you share the link

merrymercy commented 1 year ago

It is an experiment we are running currently. More details will be revealed later.

dillfrescott commented 10 months ago

When is "later"?

ntsagkas commented 9 months ago

a conversation log with this model. Truly remarkable https://docs.google.com/document/d/1_LnQx9j0yJyQ2Fwscce9lk5mNg3qpbSBv9Y9YgAfz6w/edit?usp=drivesdk

cvjxendra commented 8 months ago

Are we later yet?

dillfrescott commented 8 months ago

I think by the time they release it, it will have been passed up by another model thats already out. Try the miqu model on hugging face, its probably already better.

dillfrescott commented 8 months ago

https://huggingface.co/miqudev/miqu-1-70b

hbf731eF commented 7 months ago

I asked this question on chat.lmsys.org and valued it better than claude-3-opus, this is from a yt with Mike Green

explain: when you think about what a market maker is doing they're effectively trying to cross the bid and the ask spread in any given stock and do so repeatedly and put that extra money into their pocket in order to conduct those transactions they have to put up a certain amount of capital that facilitates them buying and selling. and the return on that capital is going to be a function of the size of the bid ask spread and the quantity of shares that are being traded . so I'm going to put Capital up against Market making activities as a function of the size of the bid ask and a function of the quantity of shares that are traded those are only the one of them the size of the bid ask is negatively correlated with size typically and the quantity of shares that gets traded is not necessarily linked to market capitalization . they tend to have higher volumes but those volumes don't scale in the same manner. right so just to put you know absolute terms on it the volume of transaction activity for apple is somewhere between five and nine times that of Delta Airlines even though its market cap is literally a 100 times larger. so those are key differences that play through that explain why we see this phenomenon but it unfortunately means that market cap weighted indices lead to the types of distortions that we're actually talking about. does that make sense yeah so if you were going to run you almost would just if you were going to run a strategy based on this to try to take advantage of it you almost would just run a market cap weighted index um rather than trying to figure out like is this stock going to be more influenced relative to to a different stock. so that is the easiest way that you do it and then the second question becomes given that I acknowledge that that is the underlying feature of the market, how do I actually beat the market and the answer unfortunately boils down to you just got to do that with leverage. so you know you've seen players when you go into periods of relative stability or relative low volatility and it becomes a unidirectional move like the easy answer is okay fine I just run with like a beta of one and a half to some underlying index that's effectively The Secret of Kathy Wood over the you know a period of time was that her beta was significantly higher than the market even as her Alpha was actually quite negative.

############################## answer by deluxe-chat-v1.2

Your explanation delves into the roles and strategies of market makers, the impact of market capitalization on trading volumes, and the implications for investment strategies. Let's break down these concepts for clarity:

Market Makers and the Bid-Ask Spread Market makers are entities (often financial institutions or firms) that facilitate trading by always being ready to buy or sell securities at publicly quoted prices. Their role is crucial in providing liquidity and reducing transaction costs for other investors. The bid-ask spread is the difference between the highest price a buyer is willing to pay (bid) and the lowest price a seller is willing to accept (ask). Market makers profit by capturing this spread, buying at the bid price, and selling at the ask price.

The return on the capital market makers deploy depends on two main factors:

The size of the bid-ask spread: Wider spreads mean higher potential profits per transaction. The volume of shares traded: More transactions can increase total profits even if the spread is narrow. Market Capitalization, Trading Volume, and Market Indices Market capitalization (market cap) is the total market value of a company's outstanding shares. It's a common measure of a company's size. However, trading volume—the quantity of shares traded in a given period—does not scale linearly with market cap. For example, a tech giant like Apple might have a market cap vastly larger than an airline like Delta, but its trading volume might only be a few times greater, not a hundredfold.

This discrepancy can lead to distortions in market cap-weighted indices, where larger companies have a more significant impact on the index's performance regardless of their trading volumes.

Investment Strategies and Beating the Market Given the understanding of market makers' operations and the relationship between market cap and trading volumes, you suggest that simply following a market cap-weighted index might seem like an easy investment strategy. However, to outperform (or "beat") the market, leveraging might be necessary. Leverage involves using borrowed money to increase the potential return of an investment.

You mention Kathy Wood, known for her investment strategies with ARK Invest, as an example of leveraging beta to outperform the market. Beta measures a security's volatility relative to the overall market; a beta greater than one suggests the security is more volatile than the market. Thus, Wood's strategy might involve higher risk (due to higher volatility) but aims for higher returns by betting on more significant movements compared to the broader market.

Conclusion In essence, understanding the mechanics of market making, the nuances of market cap versus trading volume, and the implications for index weighting can inform sophisticated investment strategies. Leveraging and targeting higher beta values can be strategies to outperform the market, but they come with increased risk. The key is balancing the desire for higher returns with the tolerance for potential losses.